Download Legal Implications of Network Economic Effects

Legal Implications of Network Economic Effects∗
Mark A. Lemley† & David McGowan‡
Economic scholarship has recently focused a great deal of attention on the phenomenon of
network externalities, or network effects: markets in which the value that consumers place on a
good increases as others buy the good. Though the economic theory of network effects is less
than fifteen years old, and is still not thoroughly understood, network effects are increasingly
playing a role in legal argument. Judges, litigators, and scholars have suggested that antitrust law,
intellectual property law, telecommunications law, Internet law, corporate law, and contract law
need to be modified to take account of network effects. Their arguments reflect a wide range of
views about what network effects are and how courts should react to them. In this Article, we explore the application of network economic theory in each of these contexts. We suggest ways in
which particular legal rules should—and should not—be modified to take account of network effects. We also attempt to draw some general conclusions about the role of network economic the-
∗
Copyright 1998 Mark A. Lemley, David McGowan, and the California Law Review, Inc.
†
Assistant Professor, University of Texas School of Law; of counsel, Fish & Richardson P.C.,
Austin, Texas
‡
Attorney, Howard, Rice, Nemerovski, Canady, Falk, & Rabkin, San Francisco, California; Lecturer,
Boalt Hall School of Law, University of California at Berkeley.
The authors would like to thank Thomas Cotter, Mel Eisenberg, Michael Froomkin, Lino Graglia,
Rose Hagan, Peter Hammer, Claire Hill, Michael Klausner, Doug Laycock, Larry Lessig, Ronald Mann, Miranda
McGowan, Peter Menell, Rob Merges, Maureen O’Rourke, Mark Patterson, David Post, Jerry Reichman, Howard
Shelanski, Peter Swire, and participants in workshops at the Boalt Hall School of Law, University of California at
Berkeley, the Chicago-Kent School of Law, the Fordham University School of Law, the Harvard Law School, the
University of Michigan Law School, and the Northwestern University School of Law for helpful comments on an
earlier draft, Shari Heino for research assistance, and the members of the cni-copyright listserv for their engaging
discussion of certain issues related to this paper.
++2
CALIFORNIA LAW REVIEW
[Vol. V:P
ory in the legal enterprise and about the way in which courts should revise legal doctrines in response to theories from fields outside the law.
I.
Parsing the Network Concept: Actual Networks, Virtual Networks, and Positive Feedback
Effects
A.Actual Networks
B.Virtual Networks
C.Positive Feedback Effects
D.Why We Should Care: The Possible “Effects” of Networks
II. Network Effects as Legal Argument
A.Antitrust
1.
United States v. Microsoft
2.
ATM Networks
3.
Credit Card Networks
4.
Standard-Setting Organizations
B.Intellectual Property
1.
Arguments for Reverse Engineering
2.
Arguments Against Protecting Interface Components
3.
Compatibility as an Antitrust Issue
4.
Conclusions
C.Government Standard-Setting
D.Telephony
E. Internet Governance
F. Corporate Law
1.
The Contractarian Paradigm
YEAR]
SHORT TITLE
2.
Potential Challenges From Network Theory
3.
Interpretive Effects
4.
Service Providers
5.
Marketing Effects
6.
Implications for Corporate Governance
3
G.Contract Terms
III. Deriving Principles for Judicial Treatment of Network Effects
A.Nature and Strength of the Effects
B.Evidence Supporting Network Effects
C.Ameliorative Effects
D.Significance of the Network Effect in the Industry
E. Importance of Existing Legal Rules
F. Nature of the Relief Sought
IV. Conclusions
As law continues to decline as an "autonomous discipline,"1 legal doctrines continue to be revised in response to theories from fields outside the law. Efforts to adapt the law to such theories
often occur in circumstances where reasoning by analogy, the bread-and-butter methodology of
litigants and courts, offers little guidance. Even where useful analogies from outside the law are
available, both the principle of stare decisis and confusion as to the role of the theory in the legal
realm may limit the rate at which courts can integrate such theories into legal doctrine. Thus, even
1.
See Richard A. Posner, The Decline of Law As an Autonomous Discipline: 1962-1987, 100 Harv. L.
Rev. 761 (1987) (referring to the loss of insularity that once characterized legal thought).
++4
CALIFORNIA LAW REVIEW
[Vol. V:P
theories relatively uncontroversial in other fields, such as the potential procompetitive effects of
vertical nonprice restraints in industrial organization theory, often arrive relatively late to the law.2
Further, disciplines outside the law may have normative conceptions of desirable outcomes,
often unacknowledged, that the law does not share. Unless the conceptual framework of the external discipline is adopted along with its theory, the law may resist adaptation even while acknowledging the soundness of the theory. Take the example of antitrust law. In most recent
antitrust disputes, economic analysis has prevailed; even the "post-Chicago" school of antitrust
analysis explicitly takes into account inefficiencies that courts long ignored, or even encouraged.
But until courts were willing to accept efficiency as a (if not always the) goal of the antitrust laws,
they resisted many economic conclusions, however evident those conclusions seemed to economists. 3 The law of corporate governance currently poses this problem. While recent theoretical
work has tended to operate from a "contractarian" paradigm, which takes as its aim the maximization of a firm's value, actual corporate statutes do not reflect such a singular focus.4
Even if there exists a reasonable consensus as to the goals of a given doctrine, the task of
adapting the theory to legal doctrine is particularly challenging where the theory is relatively new
and untested in its field of origin. In some cases, a new theory will pose questions that the law has
not previously decided. Because these novel questions are less subject to the brake of stare deci-
2.
See Continental T.V., Inc. v. GTE Sylvania, Inc., 433 U.S. 36 (1977); Business Electronics v. Sharp
Electronics, 485 U.S. 717 (1988).
3.
For a discussion of the interrelationship between theory and fact in antitrust law, see Timothy J.
Muris, Economics and Antitrust, 5 Geo. Mason L. Rev. 303 (1997).
4
The problem is compounded because governance law is state law. Despite Delaware's predominance
among large firms, varying or contradictory goals of corporate governance are common.
YEAR]
SHORT TITLE
5
sis, judicial adaptation in these areas may be more rapid. By the same token, a theory’s novelty
will limit the degree to which analogy can serve as a useful method of analysis. If no prior cases
meaningfully limit the range of possible decisions, there will be few prior cases from which to derive rules necessary to reach a decision.
In this article, we examine one important case study in the legal adaptation of a novel theory
developed outside the law. The example we choose, the relatively recent appearance in the cases
and legal commentary of references to "network externalities" or "network effects,"5 raises a number of important substantive and methodological questions. As we shall see, judicial adoption and
application of the theory of network effects has been swift in a wide variety of disciplines. Scholars have been even more ambitious than the courts, asserting arguments based on network effects
5.
Liebowitz and Margolis argue that the term “network effects” should be applied to markets with
increasing returns to scale and the term “network externalities” reserved for markets in which increasing returns create
suboptimal conditions. S.J. Liebowitz & Stephen E. Margolis, Network Externality: An Uncommon Tragedy, 8 J.
Econ. Persp. 133, 135 (1994) [hereinafter Liebowitz & Margolis, Uncommon Tragedy]. This distinction is both
appropriate and important in analyzing the intersection between economic theory and the law. The term “externality”
is generally reserved for those situations in which “an agent does not bear the full cost of his actions.” Dennis W.
Carlton & J. Mark Klamer, The Need For Coordination Among Firms, With Special Reference To Network
Industries, 50 U. Chi. L. Rev. 446, 450 n.15 (1983). Because our goal is in large part to suggest standards courts
may use to distinguish cases in which external effects should modify legal analysis from cases in which such effects
do not warrant modification of legal analysis, we generally refer to network effects, reserving the more suggestive
term “externality” for those situations we believe are likely to entail some form of inefficiency attributable to
network characteristics. As Klausner notes, the term “network externalities” seems to have occupied the entire field
of network theory. Michael Klausner, Corporations, Corporate Law, And Networks of Contracts, 81 Va. L. Rev.
757, 764 n.16 (1995).
++6
CALIFORNIA LAW REVIEW
[Vol. V:P
in a wide array of disciplines.6 At one level, these trends are laudable; to the extent network theory
succeeds in identifying issues previously overlooked, it has at least the potential to enrich whatever
legal doctrines legitimately fall within its parameters. Realizing that potential, however, will be
harder than at least some courts and commentators believe. Significant confusion remains as to
what constitutes a “network effect,” and how such effects should be used in the law.
“Network effects” refers to a group of theories clustered around the question whether and to
what extent standard economic theory must be altered in cases where "the utility that a user derives
from consumption of a good increases with the number of other agents consuming the good."7 In
other words, a network effect exists where purchasers find a good more valuable as additional purchasers buy the same good. Farrell and Saloner define the term not just to apply to circumstances
in which the goods purchased are identical, but also to include situations in which "one consumer's
value for a good increases when another consumer has a compatible good."8 We use network effects here to include both networks of like goods and goods compatible with the network. Esti-
6
For an extremely broad application of both path dependence and network effects, see Douglass C.
North, Institutions, Institutional Change and Economic Performance 7-8 (1990). And Nichoas Economides has
collected
several
hundred
published
articles
on
netowrk
effects
in
economics.
See
http://raven.stern.nyu.edu/networks/biblio_hframe.html (visited Dec. 11, 1997).
7.
Michael L. Katz & Carl Shapiro, Network Externalities, Competition, and Compatibility, 75 Am.
Econ. Rev. 424, 424 (1985) [hereinafter Katz & Shapiro, Network Externalities]. See also Philip H. Dybvig &
Chester S. Spatt, Adoption Externalities as Public Goods, 20 J. Pub. Econ. 231, 231-32 (1983).
8.
Joseph Farrell & Garth Saloner, Standardization, Compatibility, and Innovation, 16 Rand J. Econ. 70,
70 (1985). “Compatible” here means that two goods can work together, or can work with a single type of
complementary good. Software is “compatible” with an operating system if it will run on that operating system;
three-prong electrical plugs are compatible with three-prong outlets, but not with two-prong outlets.
YEAR]
SHORT TITLE
7
mates of this value added by participation in a network are quite high; “Metcalfe’s Law” asserts that
for computers, the value of participation on a network grows exponentially with the size of the
network. 9
Network effects directly challenge an important tenet of classical economic theory, which posits declining (and eventually negative) returns to scale in most markets.10 Classical theory comes
closest to addressing this problem in its treatment of natural monopolies, a designation properly
reserved for cases in which the minimum efficient size of a single producer is adequate to supply
the quantity demanded in a given market.11 But natural monopoly and network effects are not the
same thing. The problem in the case of natural monopoly is one of scale economies of supply: the
marginal and average costs of production decline throughout the demand curve for a particular
market. By contrast, network effects are demand-side rather than supply-side effects: the shape of
the demand curve is affected by existing demand. Economists long believed that natural monopo-
9.
See George Gilder, Metcalfe’s Law and Legacy, Forbes ASAP, Sept. 13, 1993, at S158.
We are
unaware of any effort to measure this effect rigorously, and we find it highly unlikely that Metcalfe’s Law in fact
holds strictly true. Nonetheless, there is clearly something to the idea that social value increases with adoption from
some goods.
10.
See, e.g., Robert S. Pindyck & Daniel L. Rubinfeld, Microeconomics 188 (2d ed. 1992)
(distinguishing among industries subject to increasing, constant, and decreasing returns to scale and noting that the
latter category “is likely to apply to any firm with large-scale operations”); David McGowan, Regulating
Competition in the Information Age: Computer Software as an Essential Facility Under the Sherman Act, 18
Hastings Comm. & Ent. L.J. 771 (1996).
11.
See, e.g., F.M Scherer, Industrial Market Structure And Economic Performance 482 (2d Ed. 1980);
Pindyck & Rubinfeld, supra note 10, at 352; McGowan, supra note 10, at 782-83.
++8
CALIFORNIA LAW REVIEW
[Vol. V:P
lies mandated governmental intervention, most particularly in the form of price regulation.12 The
case for automatic price regulation in natural monopoly markets has weakened in recent years,13 but
the analytical structure remains essentially the same. Classical theory approaches most increasing
returns problems as if they involve economies in the scale of production, rather than value added to
existing users through increased demand. While the two problems may be difficult to distinguish
in practice—and some courts may treat them the same—the two cases are analytically distinct and
therefore may require different legal treatment, as we shall see.14
That network effects exist to a greater or lesser degree in various markets is uncontroversial;
whether and to what extent existing legal theories should be revised in light of this fact is not.15
While not an autonomous discipline, neither is law merely the repository of collected abstractions
from whatever disciplines might be implicated by a particular issue. Whether through litigation or
legislation, the law must actively set a goal and devise a plan for achieving it, drawing upon whatever learning is useful in the endeavor. This requires the resolution of concrete disputes among
parties with real-world interests. From a legal perspective, therefore, the question presented by
network theories is whether we should think differently about fields such as antitrust, corporations, copyright, or contracts to the extent they seek to regulate the behavior of participants in network markets. The question must be decided as much by reference to the concerns already
embodied in the theories and doctrines of these fields as by reference to new insights generated by
network theory.
12.
See, e.g., Pindyck & Rubinfeld, supra note 10, at 352.
13.
See generally Richard J. Pierce Jr., Economic Regulation: Cases and Materials (1994).
14.
See infra notes 498-504.
15.
See infra notes 481-542 and accompanying text.
YEAR]
SHORT TITLE
9
Like most interesting questions, those presented by efforts to adapt network theory to the law
are interesting precisely because they are difficult. The theoretical implications of network markets
have not been fully elaborated even in the economic literature.16 The theoretical legal analysis that
has drawn upon such literature is even less complete, and empirical data on the behavior of firms
and consumers in network markets is scarcer still.17 Nevertheless, the law often cannot wait for
adequate theoretical development, much less firm empirical conclusions. Actual disputes among
actual parties must be resolved, and courts should be encouraged to draw upon such insights as are
available, even if they are not fully developed. Indeed, given the necessity of resolving disputes,
the law is as often a source of data as it is a product.18
In this article, we treat in an integrated fashion the adoption of network effects theories in
seven separate fields of law. Given this broad focus, our primary aim is not conclusively to resolve doctrinal disputes in each of those areas, though we do offer some views in each area based
on the present state of the evidence. Rather, our purpose in looking beyond any single set of legal
rules is to attempt to offer insight into how network effects should be treated in law, and, by impli-
16.
See Liebowitz & Margolis, Uncommon Tragedy, supra note 5, at 149 (“[T]he a priori case for
network externalities is treacherous and the empirical case is yet to be presented.”).
17.
See id.; see also Klausner, supra note 5, at 851 (suggesting that theoretical case for network effects in
corporate law is strong but that “[t]he pervasiveness and magnitude of these network externalities are empirical issues
that must be addressed in future work”).
18.
See Oliver Williamson, Book Review: The Firm, The Markets and the Law, 77 Calif. L. Rev. 223,
228 (1989) (noting that Coase drew many of his examples from judicial opinions, which recognized the reciprocal
nature of pollution before many economists did).
++10
CALIFORNIA LAW REVIEW
cation, how the law can integrate outside theories and evidence.19 Part I briefly recounts the evolution and current state of the theory of network effects in the economic literature. Part II traces the
growing use of the theory in the law, with particular emphasis on intellectual property, antitrust,
corporate law, and contract.20 Part III draws together the seemingly disparate strands of network
theory embodied in the cases and commentary, and offers some general standards for adapting
network theory to the law.
This analysis is designed to isolate the legally salient aspects of network effects theory that
exist independent of legal context and to suggest indicia that courts and legislatures may use in applying network effects theory to legal issues. Specifically, we draw the following conclusions,
elaborated in more detail in Part III:
•
Law is perhaps uniquely susceptible to fads imported from other doctrines. Because case law
is the product of an adversarial process, counsel tend to push new theories as far as possible,
regardless whether they exceed the premises of the theory, or indeed whether the theory properly applies to the case in the first place. Although opposing counsel will have incentive to
counter such arguments, it is unrealistic to expect judges, by definition busy generalists, to discern the wheat from the chaff. 21 Nor, with due respect for the significant advances in legal
scholarship as the law has become less autonomous, is the academy immune from the occa-
19.
For other attempts along these lines, see Thomas S. Ulen, Firmly Grounded: Economics in the
Future of Law, 1997 Wis. L. Rev. 433.
20.
Other legal fields in which network effects have played a role include telecommunications, Internet
law, government standard-setting, and even racial discrimination.
21
This is particularly true for trial judges, whose discretion is relatively limited and whose dockets
require more attention than appellate judges.
[Vol.
YEAR]
SHORT TITLE
11
sional tendency to take new theories to excess. Scholars may test the outer boundaries of a
theory with fewer potentially adverse consequences than courts and, if for no other reason, it is
desirable that they do. By the same token, academics’ work should be closely examined so
that courts and legislatures will have the benefit of a full dialogue. It is in this spirit that we
pursue much of what follows below. The diffusion of network economic theory into law has
been fast and furious. While law cannot and should not ignore evidence from the world
around it, there are risks associated with uncritically adopting network theory—or any other
theory—without a careful evaluation of both the theory and the facts. Some of the examples
we offer highlight this problem.
•
Network effects are complex and differentiated economic phenomena, and thus are not wellsuited to either fast or furious adaptation. It is impossible to speak meaningfully about the
“implications” of network effects without recognizing the differing factual, economic and legal
contexts in which the theory is raised. To cite only one example, certain market conditions,
such as a credible market commitment to open standards and compatibility, may ameliorate otherwise negative network effects. To the extent courts or commentators attempt to generalize
about network effects or to reason by analogy without a deep understanding of how network
economics work, they risk doing more harm than good. Several of the problems in Part II result from failure to understand network effects or their consequences. Further, some of the arguments we will consider are simply not as strong as others.
•
As with any theory from outside the law, simply understanding network effects will not necessarily, or even very often, dictate a result in a particular field of law. As with economics generally, network theory may inform courts and legislators of certain aspects of economic reality,
but how the law reacts to that reality will depend on the goals of a particular legal doctrine.
++12
CALIFORNIA LAW REVIEW
Where those goals are poorly understood, as historically was the case in antitrust,22 all the economic sophistication in the world will not prevent courts from going astray. And even where
legal goals are fully understood, they may simply be incommensurate with economic theory.
•
Finally, in those circumstances in which courts or legislatures have correctly identified problems resulting from network effects, they will often have at their disposal a variety of different
remedies. We suggest that the appropriate legal accommodation of network effects is the one
that least intrudes on existing legal doctrine. In some cases, that may mean doing nothing, especially where the network market structure is either socially optimal or self-correcting. In
other cases, to the extent possible within the confines of our adversary system, courts should
seek to solve network problems using the narrowest possible legal theory. Where a party presents a network argument in support of a very strong claim (say for violation of the antitrust
laws), and the network aspect of the problem could be solved through a more tailored approach
(say by adopting an information-forcing rule in contract law), society may be better served by
choosing the narrower doctrine and refusing to endorse a network-based theory with broader
implications than necessary.23
To explain the reasoning behind these conclusions, it is necessary first to describe the theory
of network effects, and how it applies in the legal world. It is to this endeavor that we now turn.
22.
Compare Richard A. Posner, Antitrust Law: An Economic Perspective (1976) (antitrust law is
primarily concerned with promoting social welfare), with Robert Lande, Wealth Transfers as the Original and
Primary Concern of Antitrust: The Efficiency Interpretation Challenged, 34 Hastings L.J. 67 (1982).
23.
One of us has suggested that this principle would lead to a different outcome in Eastman Kodak Co. v.
Image Technical Servs., 504 U.S. 451 (1992). See McGowan, supra note 10, at 840-41.
[Vol.
YEAR]
SHORT TITLE
13
I
Parsing The Network Concept: Actual Networks, Virtual Networks, And Positive Feedback Effects
“This is not fringe economics anymore. It’s mainstream.”
Garth Saloner24
Many things may increase in value as the number of users increases. The term network
effects therefore must be used with great care, for it encompasses a number of distinct conditions
in which value may increase with consumption. The state of both theoretical development and
empirical research varies, and the confidence with which the law uses network theory as a basis for
modifying or extending existing doctrine should be calibrated accordingly. Following Katz and
Shapiro, we view network markets as falling on a continuum that may roughly be divided into
actual networks, virtual networks, and simple positive-feedback phenomena.25 The
essential
criterion for locating a good along this continuum is the degree to which the good provides inherent
value to a consumer apart from any network characteristics. The greater the inherent value of the
good relative to any value added by additional consumers, the less significant the network effect.
24.
James Daly, The Robin Hood of the Rich, Wired, Aug. 1997, at 108, 112 (quoting economist Garth
Saloner).
25.
See Katz & Shapiro, Network Externalities, supra note 7, at 424 (distinguishing beween “direct” and
“indirect” network effects, and further discussing positive consumption externalities that may arise in markets
relating to durable goods). In a later article Katz and Shapiro adopt the term “virtual networks” for the second
category, and we follow that terminology here. See Michael L. Katz & Carl Shapiro, Systems Competition and
Network Effects, 8 J. Econ. Persp. 93, 95 (1994) [hereinafter Katz & Shapiro, Systems Competition].
++14
CALIFORNIA LAW REVIEW
A. Actual Networks
The archetypal examples of network markets involve products whose entire value lies in facilitating interactions between a consumer and others who own the product. The benefit to a purchaser, in other words, is access to other purchasers. Telephones and fax machines are classic
examples of actual network goods; owning the only telephone or fax machine in the world would
be of little benefit because it couldn’t be used to communicate with anyone. The value of the telephone or fax machine one has already purchased increases with each additional purchaser, so long
as all machines operate on the same standards and the network infrastructure is capable of processing all member communications reliably.26 In this relatively strict sense, actual networks are
effectively limited to communications markets.27 The principal characteristics distinguishing such
products from others discussed below are the absence of inherent value and the necessity for common standards among goods incorporated into the network.28
26.
See, e.g., Katz & Shapiro, Network Externalties, supra note 7, at 424 (using telephones as example
of direct network market); Klausner, supra note 5, at 772 (same); Liebowitz & Margolis, Uncommon Tragedy, supra
note 5, at 139-40 (“The paradigmatic case for a direct network effect, if not an externality, is the network of
telephone users.”).
27.
Katz and Shapiro, for example, limit their discussion of direct network effects to communications
technologies. Katz & Shapiro, Network Externalities, supra note 7, at 424-25. See infra notes 286-305 and
accompanying text (discussing the implications of network theory for telecommunications law).
Of course,
“communications technology” is not a term entirely free from doubt; for example, the Internet has some but not all
of the characteristics of a telephone network. See infra notes 306-313 and accompanying text.
28.
See Klausner, supra note 5, at 772 (noting low inherent benefit of telephone equipment); Katz &
Shapiro, Network Externalities, supra note 7, at 425 (noting importance of technological compatibility).
[Vol.
YEAR]
SHORT TITLE
15
Even these relatively strict limitations raise difficult definitional questions. Language, for example, is the fundamental medium of communication and could be said to have both negligible inherent value to the first speaker and increasing value over the range of additional speakers.
Defined as a system of symbols conveying common meanings to the proficient, certain
"languages" may resemble actual networks;29 mathematics is perhaps the clearest example. The
precision of mathematical terms may provide a close analogy to actual network connections, even
though no actual connection among "speakers" of mathematics need exist.30 Other specialized languages, such as accounting, will share at least some characteristics with actual networks, including
at a minimum a need to ensure common definitions (protocols) to facilitate communication among a
set of users. Languages are also hard to learn, which means that most people cannot simply learn
five or ten different ones to facilitate communication. Groups of people who communicate regularly may therefore converge on a single language, or at least a small number of them. As we discuss below, corporate legal scholars have raised network effects arguments that are in significant
part linguistic in form.31
Language, however, differs in significant respects from the telephone network. The differences highlight the important role legal rules of ownership play in facilitating the formation of networks. Property rights (most importantly, the right to exclude others from a network) play a
29.
See Liebowitz & Margolis, Uncommon Tragedy, supra note 5, at 136 (treating language as a case of a
“metaphorical network” in which there may be direct interaction without physical connections among network
participants).
30
Indeed, the "language" of mathematics underlies the technology employed to make many actual
networks function.
31.
See infra notes 352-463 and accompanying text.
++16
CALIFORNIA LAW REVIEW
crucial role in network markets. In many potential networks, property rights created by legal rules,
rather than physical laws, set the boundary conditions for the network. Actual networks such as
telephone lines require capital investments in physical infrastructure. Such networks therefore may
be owned: there are tangible assets to which property rights may be attached.32 Even where the
capital investment in a network is negligible, the law can establish ownership rights by fiat (in this
case by awarding exclusive rights in “intellectual property”). Where the law establishes a right to
exclude others from the use of a thing—as with intellectual property—it constrains the ability of
consumers to move between network standards, and it gives control over access and pricing to the
owner of the intellectual property embodied in the standard.
Property rights in natural languages would be difficult to conceptualize and even harder to enforce—much harder than rights in telephone lines, or even in lines of computer code.33 It is conceivable to exclude "speakers" from using more specialized languages, 34 but as a general rule the
32.
Liebowitz and Margolis make this point as well, supra note 5, at 135-36.
33.
As Robert Ellickson rightly notes in deeming language an example of nonhierarchical coordination,
“[m]illions of people have incrementally helped shape the English language into an enormously ornate and valuable
institution. Those who have contributed to this achievment have acted without the help of the state or any other
hierarchical coordinator.” Robert C. Ellickson, Order Without Law 5 (1991).
34.
The district court in Lotus Dev. Corp. v. Paperback Software, 740 F. Supp. 37 (D. Mass. 1990),
seemed to open the door to copyright protection of computer languages, such as that in the hierarchy of the Lotus 12-3 spreadsheet, but it is unlikely that that decision remains good law after the First Circuit’s holding in Lotus Dev.
Corp. v. Borland Int’l, 49 F.3d 807 (1st Cir. 1995) (holding that menu command hierarchy for the Lotus 1-2-3
spreadsheet program was uncopyrightable). A recent decision in Australia extended copyright protection to a
computer language. Data Access Corp. v. Powerflex Services Pty., No. VG473 (Fed. Ct. Aus., Melbourne 1996).
Most commentators have been critical of this approach, however. See, e.g., Pamela Samuelson, How to Interpret
[Vol.
YEAR]
SHORT TITLE
17
law does not confer exclusive rights over languages, either natural or artificial. Nor are there technical means of excluding people from using a language. Moreover, language is an extremely diverse phenomenon, subject to many cultural and geographic variables, even within the relatively
rarefied realms of law or business. Outside of mathematics, the precision with which speakers of a
given language will understand other speakers' use of the language is likely to be highly variable.
The extent of linguistic network effects, if any, will therefore be very difficult to determine.
Finally, some types of networks warrant distinct analysis because they include aspects of both
actual networks and other types of networks. The Internet, for example, presents a difficult case:
communication on the Internet requires adherence to common technical protocols that, in an
elemental sense, define the scope of the Internet.35 Internet users are not actually linked to one
another, however.
Rather, the protocols that define the “boundaries” of the Internet merely
facilitate software interoperability. Further, not everyone on the Internet wants to be connected to
everyone else; some would prefer that access to the Internet remain limited, or at least that they be
able to shut themselves off from the masses in a private group. Nonetheless, at least a significant
part of the value of the Internet itself is the fact that it enables communication among adopters; in
this sense, it is a classic actual network with characteristics similar to the telephone. We will
the Lotus Decision; And How Not To, Communications of the ACM, Nov. 1990, at 27; Marci Hamilton & Ted
Sabety, Computer Science Concepts in Copyright Cases: The Path to a Coherent Law, 10 Harv. J.L. & Tech. 239,
274-75 (1997); Elizabeth G. Lowry, Comment, Copyright Protection for Computer Languages: Creative Incentive
or Technological Threat, 39 Emory L.J. 1293, 1335 (1990). In any event, it is not clear that even those few courts
that would grant copyright protection to computer languages would afford the same protection to natural languages.
35.
See Mark A. Lemley, Antitrust and the Internet Standardization Problem, 28 Conn. L. Rev. 1041,
1043-45 (1996) [hereinafter Lemley, Internet Standardization].
++18
CALIFORNIA LAW REVIEW
discuss some aspects of the Internet in more detail below;36 we raise the issue here to point out that
networks need be neither entirely “actual” nor entirely “virtual.”
B. Virtual Networks
Goods constitute virtual networks when they provide inherent value to consumers that increases with the number of additional users of identical and/or interoperable goods. Virtual network goods need not be linked to a common system as are the constituents of a communications
network; very strong positive feedback effects tied to functional compatibility are sufficient. Computer software is the paradigm example.37 Unlike telephones and fax machines, an operating system or application program will allow even a single user to perform a variety of tasks regardless
whether even a single other consumer owns the software. At the same time, the value of a given
program grows considerably as the number of additional purchasers increases. As more consumers adopted WordPerfect, for example, it became easier for each previous user to share files without the need for a conversion program and easier for employees to switch jobs without retraining.
And as Microsoft Word has replaced WordPerfect as the word processing program of choice, it in
turn gained the benefits of widespread adoption. Data sharing in this sense requires direct horizontal technological compatibility akin to that required for telephones and fax machines to work
together, but does not require the actual connections that communications networks do. Further,
the existence of conversion software may expand the network beyond a single, proprietary product.
36.
See infra notes 306-347 and accompanying text.
37.
See, e.g., Katz & Shapiro, Network Externalities, supra note 7, at 424 (noting computer software as
example of what we here call virtual network goods); Farrell & Saloner, supra note 8, at 70 (same); Pindyck &
Rubinfeld, supra note 10, at 122 (same).
[Vol.
YEAR]
SHORT TITLE
19
In addition to horizontal technological compatibility, software be subject to “increasing returns” based on positive feedback from the market in the form of complementary goods. Software
developers will write more applications programs for an operating system with two-thirds of the
market than for a system with one-third because the operating system with the larger share will
provide the biggest market for applications programs. The availability of a broader array of application programs will reinforce the popularity of an operating system, which in turn will make investment in application programs compatible with that system more desirable than investment in
programs compatible with less popular systems.38 Similarly, firms that adopt relatively popular
software will likely incur lower costs to train employees and will find it easier to hire productive
temporary help than will firms with unpopular software. Importantly, the strength of network effects will vary depending on the type of software in question. Network effects will be materially
greater for operating systems software than for applications programs, for example,39 and a proper
legal analysis of network effects in software markets must account for this difference.
38.
See, e.g., Mark A. Lemley & David W. O’Brien, Encouraging Software Reuse, 49 Stan. L. Rev. 255,
287 (1997) (discussing network effects associated with software); McGowan, Regulating Competition, supra note
23, at 838-39 (discussing positive feedback effects relating to software); Pindyck & Rubinfeld, supra note 10, at 122
(same); Douglas G. Baird, Robert H. Gertner & Randall C. Picker, Game Theory And The Law 208-13 (1994)
(computer operating systems characterized by network effects); Katz & Shapiro, Systems Competition, supra note
25, at 99 (predicting that consumer expectations of software availability will drive choice of hardware systems).
39.
We discuss this in more detail infra notes 72-83 and accompanying text. For a good overview of the
relative roles of operating systems software and applications program software see Peter S. Menell, Tailoring Legal
Protection for Computer Software, 39 Stan. L. Rev. 1329, 1341-44 (1987). For a discussion of the relative strength
of network effects on operating systems and application programs, see Lemley & O’Brien, supra note 38, at 287-88.
++20
CALIFORNIA LAW REVIEW
Of course, technology comprises only one element of virtual networks. Like actual networks,
virtual networks are likely to require intricate webs of both formal and informal contracts to create
the value the network delivers.40 Bank-issued credit cards provide a good example.41 Although
they might confer some utility on their own (particularly in their credit aspect), credit cards exhibit
network effects because their utility increases dramatically as a network develops. As the number
of merchants willing to accept a card grows, the utility of the card to consumers increases, thus
likely increasing the number of consumers who will want to own the card, which in turn provides
incentive for more merchants to accept the card, and so on.42 With innovation in computer and
telephone technology yielding such benefits as real-time transaction processing, including such
features as fraud detection and verification of available credit,43 transactions involving bank-issued
credit cards have come to resemble interactions on an actual network.
But the technological links and potential for positive returns to scale in the credit card industry
cannot themselves create value without a sophisticated system of contracts, including agreement on
For a discussion of the potential conflicts among operating systems firms and applications firms see McGowan,
supra note 10, at 838-41 and infra notes 72-105 and accompanying text.
40.
Which is of course not to diminish the role of contract in actual networks. As we discuss below,
contract rules are central to network problems in both actual and virtual networks. Our discussion reflects the
relatively greater importance of capital investment to the creation of actual networks and the at least potentially
greater importance of contracts to the constitution of virtual networks.
41.
Bank-issued cards are to be distinguished from cards issued by retail merchants, such as a Nordstrom or
Sears, and from proprietary cards owned by a single firm, such as American Express. See Dennis W. Carlton &
Alan S. Frankel, The Antitrust Economics of Credit Card Networks, 63 Antitrust L.J. 643, 646 (1995).
42.
See infra notes 135-151 and accompanying text (discussing the economics of credit cards).
43.
See Carlton & Frankel, supra note 41, at 646.
[Vol.
YEAR]
SHORT TITLE
21
the compensation they will receive and the rules governing their conduct relative to the network.44
Thus, merchants will have a contractual relationship with a bank, which will to some extent be
subject to the bank's contractual relationship with the credit card entity. If the merchant's bank did
not issue the consumer's credit card, it in turn will have a contractual relationship with the issuing
bank pursuant to which transactions may be cleared. The issuing bank will of course have a contractual relationship with the consumer. All of these contracts are as vital to the functioning of the
credit card network as are the electronic links that facilitate transactions.
Many of these contracts are standardized by the rules of the Visa and MasterCard joint ventures. Those rules govern general network membership, such as the manner in which member
banks may use the Visa and MasterCard marks, communication among member banks, and fees
charged by member banks for processing transactions with one another. The rules do not specify
standard terms, however, for contracts between merchants and their banks or between consumers
and their banks.45 Therefore, the degree to which network theory plays a role in the legal analysis
of credit card networks depends in significant part upon the legal and economic analysis of contract
law, including the relative efficiency of standard contract terms versus either a joint venture or
horizontal integration, and limitations on the ability to contract (such as those imposed by antitrust
law).
Network markets may also implicate informal contracts—reciprocal economic relationships
that are neither sufficiently negotiated nor memorialized to be enforceable under current contract
44.
See id. at 647-48 (discussing flow of funds among participants in credit card transactions).
45.
See id. at 648.
++22
CALIFORNIA LAW REVIEW
doctrine.46 Given the powerful positive feedback effects associated with complementary goods,
particularly vertically interoperable goods such as operating systems and application program
software, such arrangements may play as important a role in forming networks as do formal
contracts. For example, a firm seeking to launch an operating system may induce applications
programmers to write compatible programs by proclaiming that the operating system will be
"open" to anyone seeking to achieve interoperability. The firm might render such pronouncements
relatively credible by such tactics as providing application programmers advance copies of new
versions of the operating system.
Even if no formal agreements are negotiated, application
programmers may dedicate significant capital to writing compatible software, creating potentially
significant benefits to consumers. Should the originator of the operating system renege once it
achieves a certain level of acceptance in the market, such informal understandings will become the
subject of significant disputes, with potentially far-reaching legal and economic implications.47
C. Positive Feedback Effects
Lastly, goods may increase in value as consumption increases even where the goods are not
themselves connections to a network and do not interoperate with like (or “compatible”) goods.
Such goods reflect little more than the need for a given degree of demand to sustain production of
the good and complementary goods and services. Where production of goods involves both fixed
and marginal costs, the average fixed costs will decline as demand for the good increases, and the
fixed costs are spread over a larger number of units. This is a common economic phenomenon—
46.
See Jeffrey N. Gordon, The Mandatory Structure of Corporate Law, 89 Colum. L. Rev. 1549, 1549
(1989) (noting economists’ “conception of a ‘contract’ as an arrangement between two or more actors supported by
reciprocal expectations and behavior”).
[Vol.
YEAR]
SHORT TITLE
23
economies of scale. In some cases, a large population may be necessary to justify any production
at all. We would intuitively expect exotic car repair shops to be more prevalent in large cities than
rural towns because a minimum concentration of car owners is required to generate sufficient demand to sustain a shop.48
Unlike actual or virtual networks, no technological compatibility, interoperability, or even
contractual relationships are necessary to sustain this "network." Strictly speaking, it is not a
network at all.
Network effects are demand-side effects—they result from the value that
consumers place on owning what other consumers already own. By contrast, economies of scale
are supply-side effects—they are a function of the cost of making the goods, and exist (at least
conceptually) regardless of consumer demand. Markets characterized by economies of scale are of
course potentially subject to material diseconomies of scale as well.
If too many consumers
purchase the same exotic car, it may become difficult to schedule repairs, obtain parts, and the like.
Similarly, once a steel plant is used to its full capacity, expanding supply will require building a
whole new plant, raising the average cost. Thus, there are definite limits in most markets to the
“value” to consumers of buying whatever other consumers want. By definition, those markets do
not exhibit network effects.
D. Why We Should Care: The Possible "Effects" of Networks
As discussed in detail and in context in the sections that follow, many of the concerns surrounding network markets are based on the presumption that such markets offer increasing returns
over a very large portion of the demand curve. Outside the realm of natural monopoly, by contrast, classical economics generally posits declining returns to scale, and thus offers few conceptual
47.
See McGowan, supra note 10, at 838-41.
++24
CALIFORNIA LAW REVIEW
tools to address the problems that arise when returns increase over a very large portion or even all
of the demand curve. Thus, arguments based on network effects may suggest that the law must
rethink the rationality of behavior considered unlikely under classical theory, such as predation in
antitrust jurisprudence, and address new risks not considered under models based on declining returns.
With respect to the behavioral issues, network markets by definition offer potentially lucrative
returns to firms that can establish their own products as standards on which competition in the
market, or in aftermarkets for complementary goods, will be based.49 This fact presents the possibility of material first-mover advantages: being the first seller in a market may confer an important
advantage over later entrants.50 Because the returns to the standards winner will be higher than in
"normal" markets, relatively risky strategies such as predation or, at a minimum, penetration pricing, might be rational in a networks market.51
Increasing returns also raise questions about the possibility of effectively leveraging a monopoly from one market to another, an argument most commonly associated with antitrust tying
claims. Chicago-school analysts have argued that leveraging is unlikely because a given amount of
48.
See Liebowitz & Margolis, Uncommon Tragedy, supra note 5, at 135.
49.
Stanley Besen & Joseph Farrell, Choosing How To Compete: Strategies and Tactics in
Standardization, 8 J. Econ. Persp. 117, 119 (1994).
50.
See id. at 122. Of course, this will not always be true. For example, the first entrant in the VCR
market (often cited as a paradigm example of network effects), the Sony Betamax, lost a standards competition to the
VHS platform.
51.
See Lemley, Internet Standardization, supra note 35, at 1074-75 (“to the extent that standardization
effects create a signficant barrier to entry in software markets once a market standard has been established, they may
make recoupment (and therefore predation) more likely”); Besen & Farrell, supra note 49, at 123.
[Vol.
YEAR]
SHORT TITLE
25
monopoly power can extract only a given amount of revenue from consumers, whether taken all in
the monopolist's primary market or split between that market and some other.52 This view has
been challenged even without regard to network theory,53 but the possibility of leveraging from a
non-network market into a network market poses an important new challenge.54 Recent activity in
the software industry also raises the possibility that markets for products that would be considered
distinct under traditional antitrust analysis,55 such as Web browsers, might simply be absorbed into
a network market through bundling with a strong network product, such as an operating system.56
52.
See, e.g., Robert H. Bork, The Antitrust Paradox: A Policy at War with Itself 372-75 (2d ed. 1993);
Posner, supra note 22, at 171-73. Indeed, even analysts outside the Chicago School have expressed skepticism for
the leverage theory of tying. See, e.g., Herbert Hovenkamp, Federal Antitrust Policy, The Law of Competition,
And Its Practice § 7.6, at 269 (1994).
53.
See Louis Kaplow, Extension of Monopoly Power Through Leveraging, 85 Colum. L. Rev. 515
(1985) (arguing that leveraging can enhance monopoly profits in cases involving market imperfections).
54.
See Lemley, Internet Standardization, supra note 35, at 1069; Bryce J. Jones II & James R. Turner,
Can an Operating System Vendor Have a Duty to Aid Its Competitors?, 37 Jurimetrics J. 355, 390 (1997)
(suggesting that leveraging is more likely in Microsoft’s OS market). Note that the argument also works in reverse:
a firm with a strong position in a network market might have enhanced ability—although perhaps less desire—to
leverage into a non-network market.
55.
The standard test is taken from Jefferson Parish Hospital District No. 2 v. Hyde, 466 U.S. 2, 21-22
(1984), under which separate markets are deemed to exist if there is sufficient consumer demand for the products to be
sold separately.
56.
See Daly, supra note 24, at 109-10 (citing Netscape lawyer Gary Reback’s claim that Microsoft is
improperly trying to drive Netscape out of the Web browser market by incorporating its competing Web browser
into new versions of its operating system). For a more detailed discussion of this allegation, see infra notes ___-___
and accompany ingtext [Professor Lemley: need to fill in].
++26
CALIFORNIA LAW REVIEW
One might also rethink unfair competition law in light of the arguably greater sensitivity of network
markets to public pronouncements: in a market where the standard product is preferred, statements
about such products might carry greater weight than in other markets.57
These arguments are closely related to the idea of "tipping," a concept Katz and Shapiro summarize as being based on
a natural tendency towards de facto standardization, which means everyone using
the same system. Because of the strong positive-feedback elements, systems markets are especially prone to ‘tipping,’ which is the tendency of one system to pull
away from its rivals in popularity once it has gained an initial edge.58
Tipping is neither inherently good nor bad. If the economics of a particular market dictate that
having one standard is more efficient than competition among systems, then "tipping" to that standard is in theory inevitable, absent significant transaction costs or some form of regulation. In
such circumstances a "tipped" market would be efficient and therefore desirable; efforts to forestall
tipping would result in suboptimal heterogeneity among systems and losses in terms of unrealized
efficiencies. That a market is best served by a single standard, however, does not always imply
that the standard should be owned by a single firm, or even that the standard should be owned at
all.
Even in markets best served by a single standard or system, however, there is at least a theoretical risk that the "wrong" standard will be adopted or that a standard that was efficient when
57.
See Katz & Shapiro, Systems Competition, supra note 25, at 107.
58.
Id. at 106.
[Vol.
YEAR]
SHORT TITLE
27
adopted will become relatively inefficient over time.59 The conclusion that a standard adopted by
consumers is suboptimal should be approached with caution.60 Setting aside for the moment the
very difficult question of deriving determinative criteria for defining “suboptimality,” consumers
might have difficulty moving to a new standard—even if they all agreed that the adopted standard
was suboptimal—because of collective action problems.
The value of any alternative system
would depend on the number of users adopting it; the rational consumer might well choose to wait
until an alternative had been adopted by others who incurred the costs of shifting to the new standard but reaped fewer benefits relative to later adopters.61
From the standpoint of legal adaptation of network theory, each of these arguments is to some
degree problematic. The presumed increasing returns of network markets are not guaranteed; networks will suffer net diseconomies of scale if the volume of interactions exceeds network capacity
and causes delays or failure.62 Positive returns to some level of scale are in any event quite com-
59.
See id. at 106 (“[S]tandardizing on a single system can be very costly if the system selected turns out
to be inferior to another system.”).
60.
See McGowan, supra note 10, at 845.
61.
In addition, there are certain more abstract concerns that have as yet played a relatively small role in
legal arguments. Principal among these is the risk of suboptimal network size, and thus of unexploited gains from
trade, stemming from the inability of existing users to compensate prospective users for the incremental value they
would add to the network by joining. Marcel Kahan and Michael Klausner have suggested in the context of corporate
bonds that intermediaries such as investment banks may ameliorate such problems to some degree. See Marcel
Kahan and Michael Klausner, Standardization and Innovation in Corporate Contracting (Or, `The Economics of
Boilerplate’), 83 Va. L. Rev. 713, 738 (1997) [hereinafter Kahan & Klausner, Standardization].
62.
One might of course posit a rational investment strategy under which the prospect of increasing
returns will induce sufficient investment to maintain the network and expand it so as to avoid such failures. For a
++28
CALIFORNIA LAW REVIEW
mon, if not ubiquitous.63 Further, network effects might not be the only effects at work. A user
might prefer Lexis to Westlaw, but only up to a certain point. If the information she needs is available only on Westlaw, she may start using that service, whatever the cost in terms of lost convenience. At a minimum, common sense tells us that there likely are differences material to most areas
of the law between a network of telephones or fax machines and a "network" of Ferrari owners. It
is thus important to analyze markets to determine the source of increasing returns—whether from
actual or virtual networks—and to distinguish among markets displaying merely positive returns to
scale, markets displaying network effects only up to a relatively low point on the demand curve, 64
and markets displaying increasing returns over most or all of the demand curve.
One final feature of network theory bears significant emphasis. Network effects tend to have
conflicting implications which are very difficult to interpret. To take corporate governance as an
example, some have argued that a given corporate governance term might display network effects
by gaining greater clarity of meaning over time and through repeated interpretation by courts. 65 If
one observes that firms all use that term, however, does that reflect maximization of positive interpretive network effects or does it reflect suboptimal tipping? Or is the term inherently the best one?
If firms use a variety of different terms on a given point, does that reflect the optimal convergence
variety of reasons, however, there is no guarantee that such a strategy would always be pursued or be properly
implemented where it was pursued. It follows that there is no guarantee that returns will invariably increase over the
full range of the demand curve even where market conditions make such a result most likely.
63.
See Liebowitz & Margolis, Uncommon Tragedy, supra note 5, at 133-35.
64.
Liebowitz & Margolis term this “inframarginal externalities.” Liebowitz & Margolis, Uncommon
Tragedy, supra note 5, at 140.
65.
Kahan & Klausner, Standardization, supra note 61, at 722.
[Vol.
YEAR]
SHORT TITLE
29
of heterogeneous firms with heterogeneous governance provisions or does it reflect opportunity
costs of not using a standard term? In many cases, the observable data can lead to diametrically
opposed conclusions, making the task of judicial adaptation extremely difficult.
The potentially inconsistent implications of network theory render the process of judicial adaptation difficult enough. That difficulty is materially compounded, however, in cases where network arguments relate to legal fields that are themselves unsettled. The economic understanding of
the potential procompetitive effects of vertical nonprice restraints entered antitrust law slowly, in
large part due to a lack of consensus among courts and commentators on the proper goals of antitrust laws. A comment by Richard Posner, writing before the Chicago revolution took very firm
hold in antitrust, applies to many fields of law today:
There are federal antitrust statutes, and they are quite brief and readable compared to
the Internal Revenue Code.
But their operative terms—"restraint of trade,"
"substantially to lessen competition," "monopolize"—are opaque; and the congressional debates and reports that preceded their enactment, and other relevant historical materials, only dimly illuminate the intended meaning of the key terms. The
courts have spent many years interpreting, or perhaps more accurately supplying,
their meaning, but the course of judicial interpretation has been so marked by contradiction and ambiguity as to leave the law in an uncertain and fluid state. What is
more, the rules of law as they are articulated and as they are applied to alter behavior are often, as is true in this instance, two quite different things.66
66
Posner, supra note 22, at 3.
++30
CALIFORNIA LAW REVIEW
This somewhat discouraged appraisal of the state of antitrust in 1976 may be less warranted
now, at least from the perspective of an economist.67 Still, many fields of law (or at least portions
of them) do resemble the state of affairs Posner described. The corporate rule that defensive measures to unwanted takeover attempts must be "reasonable in relation to the threat posed" is hardly a
paragon of clarity, nor have the copyright statutes or the cases interpreting them shed much light on
what constitutes a "fair use" of another's work.68 Such opaqueness might simply reflect a preference for ex post resolution of disputes. But in a fair number of cases, as was the case with antitrust at the time Judge Posner was writing, the law has not yet decided on a direction.
The inherent complexities and potentially inconsistent signals of network theory are difficult
but potentially surmountable problems; indecision with respect to the purpose of the law in
question is not. All law is purposive, and the process of judicial adaptation requires first a
rigorous analysis of the purposes the law seeks to achieve before the process of adapting network
theory can begin in earnest. To the extent the legal analysis is lacking, any efforts at adaptation are
67.
There is in any event greater certainty in antitrust jurisprudence today than there was twenty years ago,
in large part because of the virtually unquestioned dominance of economic tools in modern antitrust analysis. See
generally William H. Page, The Chicago School and the Evolution of Antitrust: Characterization, Antitrust Injury
and Evidentiary Sufficiency, 75 Va. L. Rev. 1221 (1989). And those economic tools have been refined over time
beyond the basic approaches characteristic (or at least sterotypic) of the Chicago School—a refinement which has
given rise to the name Post-Chicago Law and Economics. See Symposium on Post-Chicago Law and Economics,
65 Chi. Kent L. Rev. 3 (1989).
68
See James Boyle, Shamans, Software, and Spleens: Law and the Construction of the Information
Society 19 (1996) (“[I]n copyright law—to a greter extent than in most other fields of legal doctrine—there is a
routine and acknowledged breakdown of the simplifying assumptions of discourse, so that mundane issues force
lawyers, judges and policymakers to return to first principles.”).
[Vol.
YEAR]
SHORT TITLE
31
unlikely to succeed and may cause a good deal of harm. With that caveat in mind, we turn to the
instances in which network theory has been advanced as a basis for revising existing legal thought.
II
Network Effects as Legal Argument
The economic literature on network effects has been integrated quite rapidly into legal argument. Courts, litigators, and a host of commentators have used claims of network effects in an
effort to influence a wide variety of legal rules. In this section, we review both the legal literature
and the developing caselaw, focusing on a number of areas in which network effects claims seem
most prevalent. What is most startling about these asserted network effects is not simply their
number and variety, but the radically different uses to which the concept is put.
A. Antitrust
Network effects claims are at base arguments about market structure, so it is not surprising
that they have increasingly found their way into antitrust law.69 Strong network effects can fundamentally alter the way in which competition works and, therefore, the proper role of antitrust law
in a network market.70 At the same time, network effects are not present in all cases, nor even in
all cases in which they are asserted to exist. Further, in some circumstances existing antitrust doc-
69
See, e.g., Joseph Kattan, Trends in Intellectual Property Antitrust Enforcement, Antitrust, Summer
1997, at 26, 29 (“In recent years, antitrust analysis has increasingly focused on network effects.”). For a contrary
view, see Daly, supra note 24, at 108, 112 (quoting economist Garth Saloner as saying of network theory “this is
not fringe economics anymore. It’s mainstream, except in antitrust which always lags”).
70.
See Lemley, Internet Standardization, supra note 35, at 1071 (suggesting that structural antitrust
enforcement will be ineffective in the presence of true networking effects).
++32
CALIFORNIA LAW REVIEW
trine already deals effectively with network effects.71 In this section, we focus on a number of circumstances in which attorneys, judges and legal commentators have offered network effects as a
reason either to invoke antitrust law or to refrain from doing so.
1. United States v. Microsoft
Perhaps the best-known of the antitrust cases involving network effects is the Antitrust Division’s monopolization suit against Microsoft.72 After a lengthy investigation by both the Federal
Trade Commission and the Antitrust Division, the latter filed suit against Microsoft in 1994. As is
common practice, a proposed consent decree, worked out in advance by the parties accompanied
the suit. 73 The consent decree put a stop to several Microsoft practices that the Division alleged to
be anticompetitive, including the use of de facto-exclusive “per processor licenses”74 and restrictive
non-disclosure agreements with external software developers. The district court heard a rather un-
71.
See Bruce R. Snapp, Network Industries and Antitrust: A Cautionary Note, at 1, in Cutting Edge
issues in Network Industries (ABA 1996).
72.
United States v. Microsoft Corp., 159 F.R.D. 318 (D.D.C. 1995), rev’d. 56 F.3d 1448 (D.C. Cir.
1995).
73.
Id. at 321.
74.
The district court explained the potential anticompetitive effect of the per-processor license. Id. at
323. See also Kenneth C. Baseman et al., Microsoft Plays Hardball: The Use of Exclusionary Pricing and Technical
Imcompatibility to Maintain Power in markets for Operating Systems Software, 40 Antitrust Bull. 265, 267-68
(1995) (criticizing Microsoft’s per-processor license); but see Robert J. Levinson, Efficiency Lost?: The Microsoft
Consent Decree, in Malcom B. Coate & Andrew N. Kleit, The Economics of the Antitrust Process 175, 182-85
(1996) (suggesting that per-processor licenses cannot be anticompetitive, because not all companies were forced to
enter into them).
[Vol.
YEAR]
SHORT TITLE
33
usual challenge to the consent decree by several unnamed Microsoft competitors, who were upset
because the decree did not address certain kinds of allegedly anticompetitive conduct (such as
“vaporware”).75 Judge Sporkin refused to approve the consent decree because he found it too narrow.76
All of the parties seemed to acknowledge the role network effects played in Microsoft’s dominance of the operating systems market.77 The district court wrote:
75.
For more detail on the allegation that Microsoft promoted vaporware—software that did not exist—in
an effort to deter competitive entry, see Microsoft, 159 F.R.D. at 334-36; Robert Prentice, Vaporware: Imaginary
High-Tech Products and Real Antitrust Liability in a Post-Chicago World, 57 Ohio St. L.J. 1163 (1996). We will
return to the relevance of network effects to the vaporware claim infra.
76.
Microsoft, 159 F.R.D. at 333. The D.C. Circuit swiftly reversed that decision, but on jurisdictional
grounds rather than on the merits of Judge Sporkin’s antitrust concerns. United States v. Microsoft, 56 F.3d 1448
(D.C. Cir. 1995).
77
See John E. Lopatka & William H. Page, Microsoft, Monopolization, and Network Externalities:
Some Uses and Abuses of Economic Theory in Antitrust Decision Making, 40 Antitrust Bull. 317, 320-21 (1995);
Mercer H. Hartz, Comment, Dominance and Duty in the European Union: A Look Through Microsoft Windows at
the Essential Facilitiess Doctrine, 11 Emory Int’l L.Rev. 189, 210-11 (1997). Which is not to say that the issue is
entirely free from doubt among economists.
In a thought-provoking study, Timothy Bresnahan and Shane
Greenstein model the analogous move from mainframe computers to client-server environments among large
businesses. They reject the hypothesis that the move was delayed because of vendor lock-in effects, instead
highlighting the technicall difficulty of the switch. Timothy Bresnahan & Shane Greenstein, Technical Progress and
Co-invention in Computing and in the Uses of Computers, 1996 Brookings Papers on Economic Activity 1.
++34
CALIFORNIA LAW REVIEW
Microsoft, the Justice Department, and a number of competitors who oppose the
entry of the decree all agree that it is very difficult to enter the operating systems
market. There are two main reasons for this, each of which reinforces the other.
First, consumers do not want to buy PCs with an operating system that does not already have a large installed base because of their concern that there will not be a
wide range of applications software available for that operating system. The second, complementary reason why there are large barriers to entry into the operating
systems market is that [independent software developers] do not want to spend time
and money developing applications for operating systems that do not have a large
installed base. They perceive that demand for that software will be low. As a result,
[original equipment manufacturers] have little incentive to license an operating system that does not have a large installed base and include it in their PCs.78
The important issue for our purposes is what antitrust law should—or indeed could—do about this
phenomenon. One might argue that attributing Microsoft’s monopoly position in the operating
systems market largely, if not entirely, to network effects means that Microsoft has done nothing
more than compete aggressively in an industry prone to standardization.79 As we discuss below, 80 it is not at all clear that the optimal number of operating systems would be greater than one
78.
Microsoft, 159 F.R.D. at 322-23.
79.
See Lemley, Internet Standardization, supra note 35, at 1068 (“In this case the monopolist may have
some legitimate claim that its monopoly has been “thrust upon it” and is therefore not illegal under the rule of
United States v. Aluminum Co. of America,[148 F.2d 416, 429 (2d Cir. 1945)]”).
80.
See infra note 500 and accompanying text.
[Vol.
YEAR]
SHORT TITLE
35
even without network effects: the high fixed and low marginal costs of producing operating system software imply significant economies of scale, which might well produce one or only a few
standard systems even without the additional benefits derived from network effects. Even assuming that operating systems are subject to strong virtual network effects, it is hard to see what the
Antitrust Division or anyone else could do to “open the market to competition” in the long run.81
Absent governmental intervention or other barriers, the market should gravitate toward the most
efficient number of operating systems, which may well be one and certainly is not many.82 The
government’s economics expert, Nobel laureate Kenneth Arrow, made a related point in his affidavit.83
Not surprisingly, Microsoft’s competitors disagreed with this analysis, arguing that the increasing returns afforded by network effects made Microsoft more dangerous to other competitors
and, by extension, to consumers. Judge Sporkin agreed with the competitors, holding that the
81.
See Lemley, Internet Standardization, supra note 35, at 1069-72 (suggesting that even structural
antitrust relief will be only a temporary solution to a real network effects problem); Alan Murray, Antitrust Isn’t
Obsolete In a Era of High-Tech, Wall St. J., Nov. 10., 1997, at 1 (“No one, of course, is proposing breaking up
Microsoft or the Windows monopoly in order to promote innovation in operating systems; that war is over, and
Microsoft won.”). On this view, the Antitrust Division’s case was destined to be “ineffective” regardless of the relief
it sought. See Jonathan Band & Masanobu Katoh, Interfaces on Trial 45 (1995) (the ineffectiveness of the Microsoft
consent decree demonstrates the shortcomings of antitrust); cf. Baseman et al., Microsoft Plays Hardball, supra note
74, at 299 (criticizing the Antitrust Division’s reponse as ineffective).
82
Whether that operating system would necessarily be Microsoft’s is another matter.
83
See Microsoft, 159 F.R.D. at 333-34.
++36
CALIFORNIA LAW REVIEW
presence of network effects should in essence be an aggravating factor in a monopolization claim.84
Judge Sporkin discounted Arrow’s affidavit on the ground that
[i]f it is concededly difficult to open up an increasing returns market to competition
once a company has obtained a monopoly position, the Government has not shown
how prospectively prohibiting violative conduct that contributed to defendant’s
achieving its monopoly position will serve to return the market to where it should
have been absent [Microsoft’s] anticompetitive practices.85
While the Division’s argument concerned effective remedies,86 Judge Sporkin’s argument focused ex post on Microsoft’s conduct, contending that network effects made it easier for the company to keep a monopoly once acquired. The court’s reference to where the market “should have
been” is particularly telling; the court appears concerned that Microsoft does not face effective
competition. Judge Sporkin traces the causes of Microsoft’s market power to a combination of
network effects and anticompetitive conduct. His evident desire is to undo the effects of both—
“returning” the market to an idealized state of competition in which multiple companies contend
vigorously but fairly against one another and the price of goods is set at or near marginal cost. If
this was the goal, Judge Sporkin was doomed to failure even before he was cut short by the D.C.
84.
Cf. Joel Klein & Preeta Bansal, International Antitrust Enforcement in the Computer Industry, 41
Vill. L. Rev. 173, 178 (1996) (suggesting that network effects in the computer industry require that the Antitrust
Division be “especially vigilant” in preventing a de facto standard-setter from manipulating a standards competition).
85.
Microsoft, 159 F.R.D. at 334 (second emphasis added).
86.
Cf. Levinson, supra note 74, at 180 (noting that the government did not attempt to “remedy” the
network effects problem).
[Vol.
YEAR]
SHORT TITLE
37
Circuit. Courts cannot normally “undo” network effects with the tools of antitrust.87 Network effects are an inherent part of certain markets, not a “market failure” for which the law must necessarily correct. The law may need to adapt to network effects, but it should neither ignore them nor
attempt to defy them.
One might characterize Judge Sporkin’s concern in a more limited sense—as a worry about
artificial barriers to entry.88 Antitrust law is properly concerned in non-network markets with high
barriers to entry, since such barriers undermine the ability of potential competitors to discipline existing firms in a market in the short run.89 Strong network effects are themselves a barrier to entry, 90 though it is not at all clear that entry into such a market ought to be encouraged. The Antitrust
Division’s theory in Microsoft appears to be somewhat different—that in a market with weak network effects, but in which competition is still desirable, a firm might employ anticompetitive practices to further raise the barriers to entry, foreclosing useful competition that otherwise might have
occurred. This is a plausible theory in certain types of “partial network effects” markets; whether
the operating systems market belongs in that category is another matter.
87.
See Lemley, Internet Standardization, supra note 35, at 1069-72 (discussing the futility of antitrust
efforts to rid the operating systems market of network effects).
88.
Joel Klein, head of the Antitrust Division, made this point in a recent article. See Klein & Bansal,
supra note 84, at 178.
89.
See, e.g., Roger D. Blair & David L. Kaserman, Antitrust Economics 32-35 (1985).
90.
See David S. Evans & Bernard Reddy, Some Economic Aspects of Network Industries and Their
Relevance to Antitrust in the Computer Industry, [ 1 s t p a g e o f a r t i c l e ? ], 22-23 in Cutting Edge Issues in
Network Industries (ABA 1996).
++38
CALIFORNIA LAW REVIEW
The implicit debate in United States v. Microsoft over the significance of network effects is
replicated in the academic commentary on the “vaporware” issue.91 The competitive effect of vaporware—allegedly predatory preannouncements of nonexistent products—depends heavily on
network effects. As Stephan Levy has observed, lying to customers about a product ship date
erodes goodwill; a company can deceive most customers only a few times before they begin to
question all the company’s product announcements. 92 For vaporware to be worthwhile, the producer must therefore capture more from an early lie than it loses later in diminished credibility.
Network effects may offer a credible explanation for such a strategy—the consumer who is deceived into waiting for a product in a standards competition may find herself with a substantial
disincentive to switch products if the vaporware product does indeed become a market standard
when it is finally introduced. By preannouncing a product, a large company may therefore influence the outcome of a standards competition in an industry characterized by network effects.93
Absent network effects, though, it is difficult to see why anyone would be concerned about
vaporware as an antitrust issue. Repeated efforts to deceive customers might be punishable as
91
In addition to claims against Microsoft, “vaporware”-style allegations were periodically made against
IBM during its period of market dominance. See Lawrence A. Sullivan, Monopolization: Coprorate Strategy, the
IBM Cases, and the Trasnsformation of the Law, 60 Tex. L. Rev. 587 (1982).
92.
Stephan M. Levy, Should “Vaporware” be an Antitrust Concern?, 42 Antitrust Bull. 33, 34 (1997).
93.
See Prentice, supra note 75, at 1226-31. See also Lemley, Internet Standardization, supra note 35, at
1074 (“Because de facto standard-setting is a high-stakes game—the winner gets a durable monopoly for several
years, and the losers get nothing—participants in this competition may be expected to try to tip the balance in their
favor.”).
[Vol.
YEAR]
SHORT TITLE
39
fraud or deceptive advertising94 if the market doesn’t discipline the company, but it is unlikely that
deception could really lead to market power in a non-network market. For this reason, the debate
over the anticompetitive harm of vaporware in the Microsoft case seems to devolve into an empirical debate over whether network effects actually exist in that market.95 (Of course, even if vaporware turns out to be potentially anticompetitive in the operating system software market, it does not
necessarily follow that antitrust law will be particularly good at distinguishing vaporware from
honest mistakes.96)
94.
False advertising may be enjoined in an action by the Federal Trade Commission pursuant to the FTC
Act, 15 U.S.C. §§ 45, 52 pursuant to state acts modeled on the Uniform Deceptive Trade Practices Act, or in a
private action by competitors under section 43(a) of the Lanham Act. See, e.g., Johnson & Johnson * Merck
Consumer Pharmaceuticals Co. v. Smithkline Beecham Corp., 960 F.2d 294 (2d Cir. 1992); see generally Lillian
R. BeVier, Competitor Suits for False Advertising Under Section 43(a) of the Lanham Act: A Puzzle in the Law of
Deception, 78 Va. L. Rev. 1 (1992).
95.
Compare Baseman et al., supra note 74, at 295-98 (network effects prevalent in Microsoft case), with
John E. Lopatka & William H. Page, Microsoft, Monopolization, and Network Externalities: Some Uses and
Abuses of Economic Theory in Antitrust Decision Making, 40 Antitrust Bull. 317, 335-55 (1995) (doubting both
the general theory of network effects and its particular application to the Microsoft case). For empirical evidence on
the network character of software markets, see Neil Gandal, Competing Compatibility Standards and Network
Externalities in the PC Software Market, 77 Rev. Econ. & Stat. 599 (1995); Shane M. Greenstein, Did Installed
Base Give an Incumbent Any (Measurable) Advantages in Federal Computer Procurement?, 24 RAND J. Econ. 19
(1992), both of which suggest that the software market exhibits network effects.
96.
On this point, see Levy, supra note 92, at 42-43; Evans & Reddy, supra note 90, at 20-21; cf.
Prentice, supra note 75, at 1243-60 (offering a standard that attempts to distinguish innocent misrepresentations from
vaporware).
++40
CALIFORNIA LAW REVIEW
A second allegation leveled at Microsoft, albeit not by the Antitrust Division, also hinges on
network effects. In a letter to the Antitrust Division in 1996, Netscape’s outside counsel, Gary
Reback, asserted that Microsoft was paying computer manufacturers and Internet service providers
to feature Microsoft’s Internet Explorer Web browser (which is already given away for free) and to
make it more difficult to use Netscape’s Navigator browser, its largest competitor.97 Once again, at
base the argument is that Microsoft is trying to “tip” a network market in its favor during a standards competition, presumably in the hope that it will capture a lucrative market with “locked-in”
customers (or at least that it will preserve its existing power in the computer operating systems
market).98
As Mark Patterson points out, though, it would not necessarily be a bad thing for consumers
if Microsoft won a standards competition.99 If consumers would have otherwise divided into two
groups purchasing incompatible software, “predatory” conduct that induces them all to buy Microsoft’s product will in fact enhance social welfare, since all consumers will benefit from the positive
97.
The text of the letter is reprinted in Gary L. Reback, Microsoft’s Illegal Conduct: Text of Netscape’s
Letter to the U.S. Justice Department, 27:2 Antitrust L. & Econ. Rev. 97 (1996).
98.
A related—and potentially more troubling—accusation is that Microsoft is attempting to “propertize”
the TCP/IP protocol stack by bundling its particular version into Windows 95. See Willow Sheremata, Barriers to
Innovation: A Monopoly, Network Externalities, and the Speed of Innovation, 42 Antitrust Bull. __ (forthcoming
1997). One commentator worries that “the TCP/IP standard will become whatever Microsoft wants it to be. This
‘public’ standard will, in fact, become a proprietary protocol manipulated by a single vendor . . .” John Davidson,
Microsoft Endangers TCP/IP Standard, Computerworld, Jan. 15, 1996, at 37.
99.
Mark Patterson, Coercion, Deception, and Other Demand-Increasing Practices in Antitrust Law, 66
Antitrust L.J. __ (forthcoming 1997).
[Vol.
YEAR]
SHORT TITLE
41
network effects of using a single product.100 Patterson suggests that a standard-enhancing move in
a network market might be procompetitive on balance, even if it eliminates competition, since consumers of the standard product will benefit from increased adoption of the standard.101 In effect, a
new dimension of network effects must be considered for antitrust purposes—whether allegedly
anticompetitive conduct serves to help tip what would otherwise be a balanced, multi-product industry toward a monolithic standard. Far from being a bad thing, as antitrust law traditionally
would conceive it, such a market-share-enhancing move could be treated as procompetitive under
Patterson’s model—at least to the extent that the social welfare benefits would have to be balanced
against any anticompetitive effects. For example, if Microsoft’s product were inferior to a competitor’s, the cost of standardizing a suboptimal product should be considered in deciding the net
effect on social welfare.102
A related issue concerns Microsoft's increasing tendency to bundle various application programs, such as word processing software or its own Web browser, with its operating system.103
The problem here is “tying” in a standard antitrust context. Antitrust law generally permits the
combined sale of two products that a manufacturer could produce separately so long as the com-
100.
Id., draft at 87 (“To the extent, then, that Microsoft (or any other seller of a browser) can encourage
consumers to settle on its browser as a de facto standard, consumers will benefit.”).
101.
Id., draft at 87-88.
102
But if consumers can tell Microsoft’s product is inferior, they might be expected to reject it in favor of
a better product, coalescing around that alternate standrad.
It is only where we should expect the standards
competitions to be biased in some way that we should worry about selecting a suboptimal standard. See Lemley,
Lemley, Internet Standardization, supra note 35, at 1073-78.
103.
See Reback, supra note 97, at __ [recommend pincite; Boalt copy currently in binding]
(making allegations regarding Microsoft’s bundling of its Web browser with its operating system).
++42
CALIFORNIA LAW REVIEW
bined sale realizes some (albeit ill-defined) level of efficiency.104 If the operating system is itself
subject to strong network effects, then the process of expanding the boundaries of that system to
include what previously would have been considered separate products presents a potentially potent case of anticompetitive tying. Again, however, if network effects are sufficiently strong at the
application level, the efficient number of Web browsers may not be very much greater than the efficient number of operating systems.105
The Justice Department brought a claim against Microsoft in 1997 for bundling the browser
into a forthcoming operating system, styling it as a violation of the 1994 consent decree and seeking to punish Microsoft for contempt. That action was still pending at this writing. In December
1997, Judge Jackson concluded that Microsoft had not acted in contempt of the consent decree, but
nonetheless issued a preliminary injunction precluding Microsoft from conditioning the sale of any
operating system on the license and installation of its Web browser.106
104.
See Hovenkamp, supra note 52, at § 10.5, at 366-69.
105.
Interoperability between products may solve such problems to some degree, as we discuss below. See
infra notes 513-516 and accompanying text..
106
United States v. Microsoft, __ F.
Supp. __
(D.D.C. Dec. 11,
1997) (available at
http://cgi.zdnet.com/cgi-bin/printme.cgi?t=zdnn) (visited Dec. 12, 1997) [check F.Supp before g o i n g to
print]. The court concluded that the government was likely to succeed on the merits of the tying claim, noting that
“the government has shown that there exists sufficient, independent consumer demand for operating systems and
Internet access software so that it is efficient for a firm to provide those products separate, as Microsoft has
concededly done.” Id.
[Vol.
YEAR]
SHORT TITLE
43
Finally, it is possible that Sun’s Java program will radically restructure the industry, opening
it to effective interoperability between operating systems. We explore the possibility in more detail
in a forthcoming paper.107
2. ATM Networks
Claims of network effect have also played a significant role in recent years in automated teller
machine (ATM) network markets.
ATM networks exhibit what Charles Rule has called
“economies of ubiquity.”108 In other words, consumers benefit from broad acceptance of their
ATM cards. Thus, agreements expanding the network of banks that will honor a particular ATM
card benefit consumers at both the newly-added bank (who can now get cash at a host of other institutions) and at all other participating banks (who now have an additional source of cash). Indeed, consumer convenience would be enhanced to the greatest degree if every bank were on the
same ATM network.109
107
David McGowan & Mark A. Lemley, Could Java Change Everything? The Competitive Propriety of
a Proprietary Standard, 43 Antitrust Bull. ___ (forthcoming 1998).
108.
Charles F. Rule, Antitrust Analysis of Joint Ventures in the Banking Industry Evaluating Shared
ATMs, at 6, in Cutting Edge Issues in Network Industries (ABA 1996).
109.
Accord The Treasurer, Inc. v. Philadelphia Nat’l Bank, 682 F. Supp. 269, 271-72 (D.N.J.), aff’d
mem. 853 F.2d 921 (3d Cir. 1988) (“[T]he principal competitive advantage of any ATM network is the number of
ATMs utilized by the system. Financial institutions prefer a large system because it increases the potential for
interbank transactions and therefore, more profit from interchange fees. Consumers generally prefer a system with a
large number of ATMs because of the greater convenience offered by such a system. In addition, because ATM
systems entail substantial capital and operating costs, a high volume of transactions is necessary to make the
machine cost effective.”).
++44
CALIFORNIA LAW REVIEW
At the same time, ATMs do not constitute an actual network effects market. First, participation in an ATM network is not necessary to the very existence of the product. Unlike a telephone,
an ATM card remains useful even if it only allows someone to get cash at her home bank. A certain level of demand will be necessary for the bank to adopt ATM technology, but additional customers do not add value to the cards of earlier customers in the same way that additional members
of communications networks do. While an intermittent demand-side benefit may exist, in which a
bank might add further ATM machines only as it reaches various plateaus of new customers, this
effect is relatively diffuse when compared to the dynamics of actual networks. Further, expansion
of ATM networks is subject to diminishing returns because members of large ATM networks will
likely already be close to a machine that accepts their card; if that is true, the incremental convenience from a new bank joining the network is likely to be low. At some point, therefore, consumers in Texas may prefer competition between networks to the marginal value of adding a fourth
bank in, say, Guatemala to their network.110 (By contrast, the marginal value to an unaffiliated
bank of joining such a network may increase with the size of the network joined.111 )
One recent case has brought this issue into sharp relief. In Money Station v. Board of Governors of the Federal Reserve System, 112 the owner of a small ATM network challenged a merger
between two other ATM networks, including the largest such network in the United States
(“MAC”). The plaintiff argued that
110.
See Donald I. Baker, Shared ATM Networks—The Antitrust Dimension, 41 Antitrust Bull. 399, 411
(1996) (noting the tradeoff between increased consumer access to cash and competitive pricing for ATM services).
111.
See BancOne Corp., 81 Fed. Res. Bull. 492, 494 n.20 (1995) (“[B]anks tend to place a greater value
on membership in a network as its membership expands”).
112.
81 F.3d 1128 (D.C. Cir. 1996), vacated, 94 F.3d 658 (D.C. Cir. 1996).
[Vol.
YEAR]
SHORT TITLE
45
[i]f the current applications are approved, the already high barriers to entry for effective competition will be at prohibitive levels because no other network . . . will
be able to attain the critical mass of ATMs necessary to support a network that could
[compete with MAC]. . . . Once MAC locks up virtually all the ATMs in the Pennsylvania-Ohio-Kentucky area as customers of its branded network through the
transaction with Mellon and NCC, there will be an insufficient number of banks to
create a “critical mass” necessary for a new network to be formed . . .113
The plaintiff’s argument was based on the premise that competition between ATM networks was
important to consumer welfare. The Federal Reserve Board rejected this idea, relying primarily on
network theory. The Board’s decision concluded that “[n]etwork externalities, such as the economies of ubiquity, tend to promote the consolidation of regional ATM networks. . . . [T]he Board
believes that, as a result of economic and market structure conditions, regions are likely to have
one dominant ATM network.”114
The D.C. Circuit reversed the Board’s decision on procedural grounds, and also appeared to
believe that the market was subject to network effects but that those effects do not excuse a merger
that would reduce potential competition in the market for ATM networks. According to the D.C.
Circuit, the Board’s response to such effects should be to encourage larger competitors, rather than
a single dominant network.115 One need not draw this conclusion, however. Particularly if one
believes network effects in ATM markets are strong, the Circuit Court's position is inconsistent
with the reasoning supporting the Microsoft consent decree (though the issues before the court in
113.
Id. at 1131.
114.
Id. at 1133 (quoting Board decision).
++46
CALIFORNIA LAW REVIEW
that case related more to jurisdiction and statutory construction than market economics) and could
well lead to decreased efficiency. At least one commentator has argued that the existence of network effects in the ATM market means that the government ought to actively seek to abolish competition between networks, on the grounds that it is costly and will ultimately serve no purpose. 116
Ironically, legal use of the network effects idea in the ATM market originally centered on the
antitrust problems raised by entering into network arrangements at all. Antitrust law has historically been hostile to agreements between competitors;117 ATM networking certainly has this characteristic. The existence of network effects in a market is a persuasive reason not to outlaw all
such agreements between competitors, without detailed inquiry into their actual economic effects.
Commentators have argued for more lenient antitrust treatment of information-sharing where network effects are present, particularly in the ATM industry. 118 This form of network effects argument is in some sense the opposite of Judge Sporkin' in the Microsoft context: here the argument
is that network effects ought to exculpate rather than inculpate.
More recently, argument about the role of network effects in the ATM industry has focused
not on the legality of ATM networks vel non, but on the standards for access to such networks.119
115.
Id. at 1133-34 & n.7. Judge Edwards dissented and would have affirmed the Board’s conclusion.
116.
Joshua B. Konvisser, Coins, Notes and Bits: The Case for Legal Tender on the Internet, 10 Harv. J.L.
& Tech. 321, 335-37 (1997).
117.
See 15 U.S.C. § 1; Eastern States Retail Lumber Dealers Ass’n v. United States, 234 U.S. 600, 614
(1914); Lemley, Internet Standardization, supra note 35, at 1079-80 (collecting sources).
118.
See, e.g., Carlton & Klamer, supra note 5, at 447; Thomas A. Piraino, Jr., The Antitrust Analysis of
Network Joint Ventures, 47 Hastings L.J. 5, 9 (1995); Rule, supra note 108, at 1.
119
See generally United States v. Electronic Payment Services Consent Decree, 59 Fed. Reg. 44,757
(1994), discussing the settlement of the Antitrust Division’s action against the MAC ATM network.
For a
[Vol.
YEAR]
SHORT TITLE
47
Challenges to the rules of ATM networks may take one of several forms. First, a bank or other
financial institution120 that is not permitted to join an ATM network may claim it has been denied
access to an “essential facility,” or that its competitors unfairly refused to deal with it.121 While essential facility claims in antitrust are rarely successful,122 numerous states have stepped into this
market with legislation mandating nondiscriminatory access to ATMs.123 Interestingly, one commentator has objected to antitrust claims demanding nondiscriminatory access, arguing that they
reduce competition by creating a “de facto merger” of the networks that interoperate.124 While this
is true in some sense, if the market is truly characterized by network effects, such interoperability
discussion of another Antitrust Division investigation, see Margaret E. Guerin-Calvert, Current Merger Policy:
Banking and ATM Network Mergers, 41 Antitrust Bull. 289, 319-21 (1996).
120
See Competition Tribunal v. Bank of Montreal, CT-95102, Complaint at ¶ 35 (Dec. 14, 1995)
(alleging that ATM networks conspired to exclude financial institutions other than banks from participating in the
network).
121
Such claims are fairly common in a variety of markets in which groups of competitors form joint
ventures. See, e.g., Northwest Wholesale Stationers, Inc. v. Pacific Stationery & Printing, 472 U.S. 284 (1985).
For more on such claims in network industries, see Lemley, Internet Standardization, supra note 35, at 1083-86;
McGowan, supra note 10, at 836-49.
122.
See Phillip Areeda, Essential Facilities: An Epithet in Need of Limiting Principles, 58 Antitrust L.J.
841, 841 (1990); McGowan, supra note 10, at 804.
123
See D. Baker & R. Brandel, The Law of Electronic Funds Transfer Systems § 25.02[4][a], § 21.01
(1995) (surveying such laws).
124
See Baker, supra note 110, at 412-13.
++48
CALIFORNIA LAW REVIEW
between standards may not be a bad thing.125 At the least, the benefits of standardization must be
weighed against the risk that the standard chosen will be suboptimal.
Second, banks may challenge exclusivity rules in ATM networks. Most banks now participate in more than one ATM network, a fact that may significantly ameliorate the potential anticompetitive effects of network consolidation.126 But in a recent case, an antitrust plaintiff alleged that
MAC’s policy precluding members from participating in a competing ATM network effectively reduced competition in the ATM network market.127
125
Cf. McGowan, supra note 10, at 833-35 (noting that interoperability is the preferred solution to
network effects problems in the software industry).
126
However, such multiple networking presents a different antitrust problem, relating to how transactions
will be routed over multiple possible networks. If a transaction could be carried over either one of two networks,
consumers and card-issuing banks would presumably benefit from a rule that routed it over the least expensive
network (at least, the network with the smallest interchange fee). For this reason, a mandatory routing rule in an
ATM network agreement might be thought anticompetitive, since it would preclude such competition and therefore
reduce the competitive benefit of having overlapping networks in the first place. See Baker, supra note 110, at 41819.
127
Marian Bank v. Electronic Payment Services, 1997 WL 367332 (D. Del. Feb. 5, 1997) [check
before going to print on F.Supp. cite]. Visa’s PLUS network has a similar policy. In particular, Rule
points out that an exclusivity rule imposed by the dominant ATM network would discourage the development of
supplemental networks which could conceivably compete with the dominant one, and which would therefore serve to
discipline the market leader. Rule, supra note 108, at 19.
[Vol.
YEAR]
SHORT TITLE
49
Finally, members of an ATM network (or consumers) might object to fee arrangements or
purchase requirements they find onerous.128 For example, a mandatory network interchange fee
significantly reduces transaction costs because it prevents each member of the network from having
to negotiate fees for interchange with each other member. On the other hand, key players in the
network can easily use such a provision to set interchange fees at a noncompetitive level.129 There
has been antitrust litigation on the related issue of ATM network rules restricting the fees that banks
may charge ATM users. In one recent case, a district court rejected an antitrust challenge to a “no
surcharge” rule imposed by the PLUS System, concluding that such a rule was not illegal price
fixing.130
In each of these latter cases, the argument is not so much about the existence of network effects in the ATM network market as it is about the ways in which the legal evaluation of permissible conduct is informed by the existence of those effects. For example, in a non-network market,
neither exclusivity nor denial of access raises significant antitrust concern in the absence of both
market power and significant barriers to entry.131 Network effects may heighten antitrust scrutiny
over a participant’s conduct in this market—just as they did with respect to some practices in the
128
For example, in Marian Bank, supra, the plaintiff alleged that in order to join an ATM access
network, it was compelled to purchase ATM processing services from the network’s chosen vendor. See Rule, supra
note 108, at 19. See generally Karen L. Grimm & David A. Balto, Consumer Pricing for ATM Services: Antitrust
Constraints and Legislative Alternatives, 9 Ga. St. U. L. Rev. 839 (1993).
129.
See Rule, supra note 108, at 17-18 (weighing pro-and anti-competitive effects of such mandatory fee
structures).
130.
Southtrust Corp. v. PLUS System, 913 F. Supp. 1517, 1519 (N.D. Ala. 1995).
131.
Blair & Kaserman, supra note 89, at 32.
++50
CALIFORNIA LAW REVIEW
operating systems market—because of concerns over increased barriers to entry.132 On the other
hand, economies of scale might suggest the existence of procompetitive efficiencies, at least in
merger cases. 133 Thus, courts and regulators must be careful to distinguish between network effects and economies of scale, since the two may have very different implications for merger analysis.134
3. Credit Card Networks
The economics of credit card networks resemble those of ATMs in many respects.135 As noted
above, once a network overcomes the chicken-and-egg problem of forming an enterprise dependent in part on existing members, consumers benefit from broad acceptance of their credit card in
much the same way that they do from broad acceptance of their ATM card. Merchants are more
likely to accept a card if everyone wants to use it, suggesting that additional consumer adoptions of
a particular card redound to the benefit of all card holders. In addition, some (but not all) credit
132.
See Baker, supra note 110, at 408 (arguing that “[n]ew entry into the ‘branded ATM network’ market
has been virtually nonexistent anywhere in the country. It requires a critical mass of cards and ATMs.”).
133.
See Hovenkamp, supra note 52, § 12.2b & c, at 452-55. It should be noted that the “efficiencies
defense” in merger cases has met with only mixed success, largely because the legislative history to the 1950 CellerKefauver Amendments to section 7 of the Clayton Act seems unambiguously to reject such a defense. See Derek
Bok, Section 7 of the Clayton Act and the Merging of Law and Economics, 74 Harv. L. Rev. 226, 234 (1960).
134.
See Guerin-Calvert, supra note 119, at 320-21 (noting Federal Reserve Board approval of an ATM
network merger, in part because of the economies of scale resulting from the merger).
135.
49.
For a discussion of how credit card transactions work, see Carlton & Frankel, supra note 41, at 645-
[Vol.
YEAR]
SHORT TITLE
51
cards are composed of networks of card-issuing banks, which also benefit from expansions in the
scope of the network since they make the card more attractive to consumers.
Nonetheless, there are subtle differences between credit card “networks”136 and ATM networks. First, unlike ATM networks, the value to consumers of adding marginal merchants does
not decline. In large part, this is because an ATM card is generally used for obtaining cash, while
a credit card is an adjunct to a financial transaction with a third party.137 Consumers need only one
convenient ATM in each location, but would prefer that every merchant with whom they do business accept their credit cards. Second, while some ATM networks are exclusive, credit card networks are nonexclusive. Consumers can and do hold more than one type of credit card, and
merchants can and do accept more than one type of card.138 This nonexclusivity means that network effects will not necessarily drive the industry toward standardization, but rather toward
overlapping credit card networks that may still compete in certain ways.
136.
Credit cards are themselves networks of credit provision, while ATM cards are generally thought of as
existing independently of ATM networks. To some extent, however, this is merely historical accident. Consider
department store or gasoline company credit cards, for example. Just as an ATM card originally provided access to
cash only at the issuing bank, so these credit cards provided access to credit only at the sponsoring store or chain of
stores. A cross-merchant credit card like American Express is an innovation analogous to the development of ATM
networks. For more on this history, see SCFC ILC, Inc. v. Visa U.S.A., Inc., 819 F. Supp. 956, 963 & n.2 (D.
Utah 1993), aff’d in part and rev’d in part, 36 F.3d 958 (10th Cir. 1994).
137.
ATM cards are increasingly used as payment options in point of sale transactions, for example at gas
stations and grocery stores. To the extent this holds true, the ATM point of sale network resembles the credit card
network more closely than it does the ATM cash access network.
138
Consumers could hold more than one ATM card, if they used more than one bank for everyday cash
needs. But most don’t.
++52
CALIFORNIA LAW REVIEW
One aspect of this nonexclusivity was at issue in SCFC ILC v. Visa USA. 139 In that case
Sears, the owner of the Discover credit card, sought to issue Visa cards as well.140 Visa denied
Sears’ application, enacting a Visa network bylaw that prevented member institutions from issuing
competing credit cards and singling out Discover cards by name.141 Sears challenged the bylaw on
antitrust grounds. The court treated the case as involving a “joint venture” (the Visa network), and
rejected per se analysis as inappropriate for such a venture.142 Rather remarkably, the court concluded that the relevant market was not the market for credit card networks, but rather the market
for card issuers (of which there were 6,000 in the Visa network alone). Since Visa did not participate in that market, the court concluded that it could not have market power.143 In so doing, the
139.
36 F.3d 958 (10th Cir. 1994), cert. denied, 515 U.S. 1152 (1995).
140.
Sears, which originally owned the Discover card before selling the business to Dean Witter, applied
for Visa membership and was turned down. Undaunted, Sears purchased the assets (including membership in the
Visa joint venture) of a failed Utah thrift from the Resolution Trust Corporation and attempted to launch a new brand
of Visa card through that entity. SCFC ILC, 819 F. Supp. at 964.
141.
Visa members were permitted to issue MasterCards, Visa’s largest credit card network competitor. 36
F.3d at 961.
142.
Id. at 963-65. The per se rule allows courts to condemn some agreements under the antitrust laws
without detailed inquiry into their market effect.
143.
Id. at 965-69. The court’s opinion is hardly a model of clarity on this point. Further, as Carlton and
Frankel observe, it simply misunderstands the nature of the market power inquiry to fail to distinguish between
decisions made individually by 6,000 banks and decisions made collectively by the Visa organization. See Carlton
& Frankel, supra note 41, at 653-54; accord Dennis W. Carlton & Steven C. Salop, You Keep on Knocking but
You Can’t Come In: Evaluating Restrictions on Access to Input Joint Ventures, 9 Harv. J.L. & Tech. 319, 337-40
(1996).
[Vol.
YEAR]
SHORT TITLE
53
court sidestepped the fundamental nature of Sears’ claim—that the Visa network was an essential
facility to which it should be allowed access on the same terms as any other card issuer because the
facility had strong network effects.144 The court did so in part because it did not think the credit
card network was so “extraordinary” as to warrant application of the essential facilities rule.145 In
addition, though, the court appeared to accept Visa’s argument that denying membership to Sears
would promote healthy “intersystem” competition by maintaining several different networks that
could compete with each other.146 The latter argument is implausible if the industry really is characterized by network effects of any strength.147 There would seem to be no greater warrant for antitrust to promote exclusive, competitive networks in such an industry than to attempt to break up
Microsoft unless the network effects were sufficiently weaker in the credit card industry to make
intersystem competition a socially optimal alternative.
As the Discover card litigation demonstrates, suits over access to the credit card network system raise many of the same sorts of group boycott and essential facilities claims discussed earlier.148
There are also some of the same fee-setting issues we saw in the ATM network context. Once
again, the role network effects should play in antitrust inquiry is unclear.149 One is left with a co-
144.
36 F.3d at 971.
145.
Id. at 971.
146.
Id. at 969.
147.
Perhaps we are too cynical, but we also find it difficult to believe that Visa’s motivation in excluding
Sears from membership its network was the pure-hearted one of wanting to encourage Sears to be a more effective
competitor of the Visa network. See id.
148.
See supra notes 112-134 and accompanying text.
149.
See, e.g., Carlton & Frankel, supra note 41, at 661 n.42 (network effects may mean that credit card
interchange fee system has procompetitive consequences).
++54
CALIFORNIA LAW REVIEW
nundrum: an entity that allows over 6,000 members to compete with one another on price (interest
rates) and on other terms does not at first glance appear anticompetitive. At the same time, it is difficult to see what harm to consumers would come from allowing competing issuers such as Discover into the system; a 6,000-member joint venture is likely not deriving a significant portion of
its value from exclusion of competitors. On the other hand, with 6,000 banks competing within
the system, allowing new members such as Sears to join might not confer much in the way of
marginal benefits to consumers.
We are left with an intuition, though admittedly little in the way of evidence or analytical
proof,150 that, if relatively strong network effects can be shown to exist, open standards are likely
to do less harm in this area than mandating competing closed systems. Remedies for achieving that
end are, admittedly, problematic. For now, we note that among other things one must take into
account the value that members of the existing network derive from the network's brand—the
"Visa" or "MasterCard" trademarks—and whether allowing other branded cards to join the network would diminish that value.151 There is little question that these marks have value, as do competitive marks such as American Express and Discover. If Discover were allowed to join the Visa
network, it might well receive benefits created by Visa's investments in brand loyalty. But that
issue could perhaps be treated in negotiations over an access price rather than by maintaining fragmentary standards in a strong network market.
150.
It is worth noting, however, that when AT&T entered the market with a no-annual-fee credit card,
most other card issuers quickly followed suit. See id. at 653.
151.
The authors wish to disclose that one of us (McGowan) has represented American Express in
trademark litigation adverse to Visa.
[Vol.
YEAR]
SHORT TITLE
55
4. Standard-Setting Organizations152
The problems with "winning" and "losing" standards in network effects industries generally arise when standards are incompatible, forcing consumers and developers to choose between
them. Competitions between incompatible standards have a "winner-take-all" quality, rewarding
the developer of a new standard by giving him the full value of the standard itself and not just the
value of his contribution.153 This result is inefficient. Network theory, which posits incremental
benefits to existing users from network growth, suggests that network goods optimally should be
priced as low as possible to allow widespread adoption of the standard.154 A proprietary standard
owned by a single, “winning” company that can set whatever price it wants is not the best way to
achieve this goal, because monopolists maximize their revenue by raising prices above the com-
152.
This section builds on the discussion in Lemley, Internet Standardization, supra note 35, at 1059-65.
153.
See David Friedman, Standards as Intellectual Property: An Economic Approach, 19 U. Dayton L.
Rev. 1109, 1121 (1994). Of course, in practice most standards competitions are not won absolutely, and minority
products (like the Apple operating system) do continue to exist. Nonetheless, it is clear that the benefits to
becoming the dominant standard-setter are substantial. But cf. Daniel J. Gifford, Microsoft Corporation, the Justice
Department, and Antitrust Theory, 25 Sw. U. L. Rev. 621, 642 (1996) (suggesting that this fringe competition
makes the network externalities argument less important).
154.
See Joseph Farrell, Arguments for Weaker Intellectual Property Protection in Network Industries,
Standards for Information Infrastructure 378, 384-89 (Brian Kahin & Janet Abbate eds., 1995) (“[W]hen network
effects are present, each user deterred by a price above cost not only loses the benefit of the innovation for himself
but also reduces the benefit to all those who do adopt. So a given reduction in demand is more harmful when
network effects are important.”).
++56
CALIFORNIA LAW REVIEW
petitive level and therefore excluding some participants (at least absent perfect price discrimination).155
One possible solution to the standardization problem is to make the competing standards interoperable. If people can switch back and forth between competing versions of what is essentially
the same standard, perhaps society can capture the benefits of competition without wasteful duplication of effort, and without stranding consumers who make the wrong choice.156 Elsewhere, we
have suggested ways in which interoperability might be achieved in the Internet context.157
155.
Nor does a proprietary standard seem necessary to encourage the production of future works of
intellectual property. While this is the purpose behind providing intellectual property protection, the winners of
standards competitions receive a windfall that is far greater than what intellectual property normally gives as an
incentive to invention.
156.
Such group standard-setting is common in the computer hardware industry. Another more recent
example is the market for digital video discs, where competing groups (Sony-Philips and Toshiba-Matsushita)
offering incompatible standards agreed to use a single compatible format incorporating elements from both products.
See My Way or the Highway, N.Y. Times, Mar. 18, 1996, at C5. There is some recent empirical evidence to
suggest that open, group-set standards produce cost savings over proprietary standards. See Joseph Bailey et al., The
Economics of Advanced Services in an Open Communications Infrastructure: Transaction Costs, Production Costs,
and Network Externalities, 4 Info. Infrastructure & Pol’y 225, 271-72 (1995); E. Robert Yoches & Kenneth M.
Frankel, Legal Implications of Standards in the Computer and Software Industries, in Intellectual Property Antitrust
[recommend pincite] (P.L.I. 1995) (“As the computer industry reaches maturity, it has learned few lessons as
important as the need for ‘open systems.’”).
157.
See Lemley, Internet Standardization, supra note 35, at 1059-65.
[Vol.
YEAR]
SHORT TITLE
57
One approach to achieving interoperable standards is for a private industry organization open
to all members to adopt a single standard.158 If the members of such a group collectively have a
significant market share, their adoption of a standard may produce the "tipping" effect described
above, bringing the rest of the industry into line.159 Private standard-setting organizations are more
efficient than government organizations in several respects.160 Because they are more marketoriented, they are less likely than their government counterparts to settle on an inefficient standard.
If they do choose an inefficient standard, it may be less entrenched than an equivalent government
158.
Such private organizations are relatively common in a variety of industries. For a historical review of
standard setting in the United States, see D. Linda Garcia, Standard Setting in the United States: Public and Private
Sector Roles, IEEE Micro, Dec. 1993, at 28-35.
Significantly, Farrell concludes for a variety of reasons that strong intellectual property protection is
likely to hamper formal private standard-setting. See Farrell, Weaker Intellectual Property, supra note 154, at 44-45;
see also Mark Shurmer & Gary Lea, Telecommunications Standardization and Intellectual Property Rights: A
Fundamental Dilemma?, in Standards for Information Infrastructure, supra note 154, at 384-89; Yoches & Frankel,
supra note 156, at __ [recommend pincite].
For more on this approach, see infra notes 191-267 and
accompanying text.
159.
Of course, not all standard-setting groups have such market control. As Libicki observes, many of the
most successful group standards started small and grew to become dominant. See Martin C. Libicki, Standards: The
Rough Road to the Common Byte, in Standards for Information Infrastructure, supra note 154, at 75; see also Jim
Isaak, Information Infrastructure Meta-Architecture and Cross-Industry Standardization, in Standards for Information
Infrastructure, supra note 158, at 100, 101 (arguing that group or open standards “must also reach the status of being
‘de facto’ to be sufficient”).
160.
For a general discussion of the benefits of group standard-setting, see Douglas D. Leeds, Raising the
Standard: Antitrust Scrutiny of Standard-Setting Consortia in High Technology Industries, 7 Fordham Intell. Prop.
Media & Ent. L.J. 641, 643-649 (1997).
++58
CALIFORNIA LAW REVIEW
standard, because private standards are potentially subject to "leapfrogging"—that is, being surpassed rapidly by a new standard. Significantly, private group standard-setting may also be more
efficient than de facto standardization, since having multiple companies participating in a standard
means that those companies can compete to offer products incorporating the standard after it is selected, thus expanding output and lowering prices.161 In Katz and Shapiro’s model, group standard-setting trades off first-round competition (to set the de facto standard) to achieve competition
within the standard in later periods.162
To be sure, private group standard-setting has problems. Private groups are not immune from
capture, particularly where the standard to be set excludes rather than includes competitors.163
Agreement on private standards can be a time-consuming process, delaying innovation based on
161.
See David J. Teece, When is Virtual Virtuous? Organizing for Innovation, Harv. Bus. Rev., Jan.-Feb.
1996, at 69 (offering the IBM PC standard as an example of an open standard that resulted in significant intrastandard
competition).
162.
See Katz & Shapiro, Systems Competition, supra note 25, at 111; William E. Cohen, Competition
and Foreclosure in the Context of Installed Base and Compatibility Effects, 64 Antitrust L.J. 535, 550 (1996).
163.
A clear example of an attempt to capture a private standard-setting organization is chronicled in Allied
Tube & Conduit Corp. v. Indian Head, Inc., 486 U.S. 492 (1988). In that case, Indian Head sought permission from
the National Fire Protection Association to market polyvinyl electrical conduit. This required the NFPA to alter the
National Electrical Code it had previously written, which allowed only steel conduit. Allied Tube, the nation’s
largest maker of steel conduit, packed the NFPA meeting and voted down the proposal. Id. at 496-97. See David
McGowan & Mark A. Lemley, Antitrust Immunity: State Action and Federalism, Petitioning and the First
Amendment, 17 Harv. J.L. & Pub. Pol’y 293, 308-11 (1994) (discussing the Allied Tube case in detail).
[Vol.
YEAR]
SHORT TITLE
59
the standard while it is being set.164 Private groups may also be a front for a cartel. Further, even
an efficient open standard may become problematic over time. While an open standard allows an
industry to evolve to improved technology as innovation progresses, that evolution may result in a
group standard that becomes more cumbersome over time as successive refinements are added to
it. For example, it is probable that Microsoft's Windows operating system is more cumbersome
than it would be if Microsoft did not have to ensure that it was compatible with early versions of its
older DOS operating system. Over time, group standards may bog down in accumulated problems
resulting from the need for backward and horizontal compatibility. Nonetheless, if the choice is
between a standard that evolves (as group standards generally do) and one that stagnates (as both
government and private de facto standards are likely to do), most consumers would choose the dynamic standard, at least in an innovation-driven market like computer or Internet software.165
Open standards present certain antitrust problems, however. The first such problem is that
standard-setting organizations may be vulnerable to challenge because they involve a horizontal
agreement between competitors in violation of section 1 of the Sherman Act. 166 While such horizontal agreements have historically been treated as illegal per se, there is reason to be optimistic that
courts will apply the more lenient rule of reason in standard-setting cases and will recognize that
164.
On this problem, and some ways to avoid it, see Joseph Farrell, Choosing the Rules for Formal
Standardization [recommend pincite] (working paper April 1993).
165.
See Shane Greenstein, Markets, Standards, and the Information Infrastructure, IEEE Micro, Dec. 1993,
at 36 (arguing that on balance the benefits of private standard-setting outweigh the costs).
166.
For more detail on this issue, see Lemley, Internet Standardization, supra note 35, at 1079-83.
++60
CALIFORNIA LAW REVIEW
the efficiencies that flow from open standard-setting in network effects industries are different and
much greater than those in ordinary industries.167
The second problem arises from the important difference between open and closed standardsetting organizations. If membership in a standard-setting organization is limited to a subset of the
full industry, and if the standard it produces is both one from which non-members can be excluded
and which possesses durable market power due in part to network effects, refusal to allow open
participation in the standard may have anticompetitive effects that antitrust law properly should address. Similar problems can occur even in nominally open groups, if voting or participation rules
are structured in such a way that a subset has effective control.
Antitrust treats such claims of exclusion from private groups in one of two ways. First, closing the group might be viewed as a horizontal group boycott or concerted refusal to deal with competitors. While the parameters of the antitrust prohibition against group boycotts are far from
167.
See Consolidated Metal Prods., Inc. v. American Petroleum Inst., 846 F.2d 284, 292 (5th Cir. 1988);
Lemley, Internet Standardization, supra note 35, at 1079-83 (citing cases and commentators); Dennis W. Carlton et
al., Communication Among Competitors: Game Theory and Antitrust, 5 Geo. Mason L. Rev. 423, 424-25
(1997); Thomas M. Jorde & David J. Teece, Rule of Reason Analysis of Horizontal Arrangements: Agreements
Designed to Advance Innovation and Commercialize Technology, 61 Antitrust L.J. 579, 600 (1993); David J. Teece,
Information Sharing, Innovation and Antitrust, 62 Antitrust L.J. 465, 475 (1994) (“The advantages to society
associated with the wide-spread adoption of common standards can be very large, as network externalities are often
considerable.”); William H. Pratt et al., Refusals to Deal in the Context of Network Joint Ventures, 52 Bus. Law.
531, 536-37 (1997); Evans & Reddy, supra note 90, at 11.
[Vol.
YEAR]
SHORT TITLE
61
clear,168 the Supreme Court's decision in Northwest Wholesale is instructive. There, the plaintiff
sued a wholesale purchasing cooperative that had denied it membership (and accompanying discounts on products purchased in bulk by the cooperative). The Court nominally applied a per se
rule condemning the joint refusal to deal, but in fact engaged in a market-power-based inquiry,
seeking to determine the importance of membership to effective competition and whether "the boycotting firms possessed a dominant position in the relevant market."169 The rule against group boycotts has also been applied (again under the rule of reason) to the New York Stock Exchange, a
body that is at least in part a standard-setting organization.170 Alternatively, antitrust might treat access to a standard-setting organization (or at least its interface standards) as an "essential facility."
Under this doctrine, the owners of facilities that are essential to effective competition must make
them available to competitors on nondiscriminatory terms. Thus, the railroads that collectively
owned the only railroad switching yard in St. Louis at the height of the railroad era were required
to give all railroads access to the yard on equal terms.171 Similarly, the essential facilities doctrine
has been used to compel regulated local telephone monopolies to interconnect all long-distance car-
168.
See, e.g., Northwest Wholesale Stationers, Inc. v. Pacific Stationery & Printing Co., 472 U.S. 284,
294 (1985). For other statements of the confusion surrounding group boycott law, see Philip Areeda, Antitrust
Analysis 381 (2d ed. 1974); Lawrence A. Sullivan, Handbook of the Law of Antitrust 229-230 (1977).
169.
Northwest Wholesale, 472 U.S. at 294. Cf. FTC v. Indiana Federation of Dentists, 476 U.S. 447,
459 (1986) (applying a “quick look” rule of reason to an agreement by dentists to deny information to insurers). The
Court in Indiana Federation held the agreement unlawful because the dentists failed to offer a procompetitive
justification for their actions. See id. at 459.
170.
See Silver v. New York Stock Exchange, 373 U.S. 341, 354-55 (1963) (refusing to invalidate NYSE
restrictions on membership, in part because the Exchange was already subject to heavy SEC regulation).
++62
CALIFORNIA LAW REVIEW
riers on substantially equal terms.172 A claim that membership in a standard-setting organization (or
at least access to its work product) was essential to competition in a networked industry arguably
would guarantee a level playing field for all competitors, though such a result would not necessarily or automatically enhance social welfare.173
Of the two theories, we are inclined to favor the former in standard-setting cases. The essential facilities doctrine has been roundly criticized as overbroad. Professor Areeda called it "an epithet in need of limiting principles."174 And the vast majority of essential facilities claims are rejected
by the courts, even in circumstances where control over a facility confers a substantial advantage
upon a competitor. 175 Further, a decision to apply the essential facilities doctrine to standardized
171.
United States v. Terminal Railroad Ass’n, 224 U.S. 383, 411 (1912).
172.
MCI v. AT&T, 708 F.2d 1081, 1132-33 (7th Cir. 1982); see Mark C. Rosenblum, The Antitrust
Rationale for the MFJ’s Line-of-Business Restrictions and a Policy Proposal for Removing Them, 25 Sw. U. L.
Rev. 605, 608-11 (1996).
173
See, e.g., supra note 163.
174.
Phillip Areeda, Essential Facilities: An Epithet in Need of Limiting Principles, 58 Antitrust L.J.
841; accord McGowan, supra note 10, at 781-82 (arguing that essential facilities claims should be limited to natural
monopoly situations); Christopher M. Seelen, Comment, The Essential Facilities Doctrine: What Does it Mean to
be Essential?, 80 Marquette L. Rev. 1117, 1123-25 (1997) (distinguishing between facilities essential to the public
and those merely essential to competitors). But see Carlton & Salop, supra note 143, at 320-21; Farrell, supra note
164, at 8-9 (arguing that essential facilities claims may be justified in the circumstances of networked markets).
175.
See, e.g., City of Anaheim v. Southern California Edison Co., 955 F.2d 1373 (9th Cir. 1992) (access
to electric power transmission lines not essential); Alaska Airlines v. United Airlines, 948 F.2d 536 (9th Cir. 1991)
(access to airline computer reservation system not essential); Illinois ex rel Burris v. Panhandle Eastern Pipe Line
Co., 935 F.2d 1469 (7th Cir. 1991) (natural gas pipeline facilities not essential); 3 Julian O. von Kalinowski,
[Vol.
YEAR]
SHORT TITLE
63
industries would offer no way to distinguish group standards from individual standards, and therefore could dramatically expand the scope of antitrust intervention in the market.176 Group boycott
claims, by contrast, attempt to distinguish concerted action to boycott a competitor (which is subject to scrutiny under Section 1 of the Sherman Act)177 from unilateral refusals to deal (which are
generally legal).178 However, recent cases such as Northwest Wholesale suggest that the law of
group boycotts is converging with the rule on essential facilities and that a group boycott will not
amount to a Section 1 violation unless the plaintiff has been denied effective access to the market.179
While the issue is not free from doubt, the use of antitrust doctrine—whether it be group boycott or essential facilities—to compel access to a standard-setting organization should probably be
Antitrust Laws and Trade Regulation § 19.05[3], at 19-124 (2d ed. 1995) (cataloging essential facilities cases). On
the other hand, the European Union may be more willing to apply the essential facilities doctrine in a broad range of
circumstances. See Harz, supra note 77, at 225-26, 230-31.
176.
But cf. Maureen A. O’Rourke, Drawing the Boundary Between Copyright and Contract: Copyright
Preemption of Software License Terms, 45 Duke L.J. 479, 547 (1995) (arguing for application of essential facilities
rule in the software industry whenever a software developer has market power); Jones & Turner, supra note 54, at
386 (“The essential facilities doctrine is easily applied to Microsoft’s possession of the leading operating system.”).
177.
See, e.g., Klors Inc. v. Broadway-Hale Stores, 359 U.S. 207, 210-12 (1959).
178.
See, e.g., Data General Corp. v. Grumman Support Servs., 36 F.3d 1147, 1182-84 (1st Cir. 1994).
Some unilateral refusals to deal have been found to violate the antitrust laws under section 2. See Aspen Skiing Co.
v. Aspen Highlands Skiing Corp., 472 U.S. 585, 610-11 (1985).
179.
See, e.g., Pratt et al., supra note 167, at 544-45 (treating the issue as involving refusals to deal); cf.
Robert Heidt, Self-Regulation and the Useless Concept “Group Boycott,” 39 Vand. L. Rev. 1507, 514-15 (1986)
(“group boycott” concept is unnecessary in modern antitrust law).
++64
CALIFORNIA LAW REVIEW
rare.180 Not every organization that attempts to set industry standards must be open to all members.181 Further, vigorous application of the antitrust laws to require access to standards groups
may discourage group standard-setting altogether, since companies may be unwilling for a variety
of reasons to discuss their product plans with certain competitors.182 This does not mean that standard-setting organizations should never be forced to open their doors, however. While there are
good reasons to limit the use of the essential facilities doctrine (which focuses on scale as such
rather than exclusionary conduct), cases involving exclusionary conduct by competitors or a single
firm seeking to obtain control of a highly standards-driven market, may well warrant application of
180.
One possible approach to take in order to limit such claims is to allow government but not private
suits for violations of the antitrust laws by standard-setting organizations. For a more general proposal along these
lines, see Edward A. Snyder & Thomas E. Kauper, Misuse of the Antitrust Laws: The Competitor Plaintiff, 90
Mich. L. Rev. 551 (1991).
181.
See National Ass’n of Review Appraisers & Mortgage Underwriters v. Appraisal Found., 64 F.3d
1130, 1137 (8th Cir. 1995).
182.
See James B. Kobak, Jr., Enforcers Focus on IP Issues, Nat’l L.J., May 6, 1996, at B7, B10. One
way to avoid this result would be to compel competitor access only to the standard itself and not to the group
producing the standard. See generally Joel R. Reidenberg, Governing Networks and Rule-Making in Cyberspace, 45
Emory L.J. 911 (1996). This approach would have the advantage of requiring less government intrusion into the
workings of the private group. Whether it would be effective, however, may depend on the complexity of the
standards at issue and the potential for group members to obtain a lead-time advantage before the standard is
announced.
[Vol.
YEAR]
SHORT TITLE
65
some antitrust rule, at least where the conduct appears likely to secure control of a market standard. 183
In evaluating claims of anticompetitive exclusion from a standard-setting organization, the
courts must take account of legitimate reasons proffered for the exclusion.184 There are unquestionably valid reasons to exclude a competitor from a group standard-setting organization, particularly if it is engaged in free riding on the cooperative efforts of the group 185 or if it is unwilling to
meet the reasonable technical or investment requirements of the group.186 On the other hand, restrictions based purely on the number or size of participants in a group seem more likely to mask a
cartel, particularly in a network effects market where more participation is generally better from a
societal perspective. And restrictions of the sort at issue in SCFC ILC v. Visa, which exclude particular competitors by name while allowing others, should be subject to the most exacting scrutiny.
183
The FTC seems to agree with this approach. See Anticipating the 21st Centruy: Competition Policy
in the New, High-Tech Gloabl Marketplace, ch. 9, at 2-3 (1996).
184.
See generally Pratt et al., supra note 167, at 553-57 (cataloging types of exclusionary bylaws).
185.
Cf. Rothery Storage & Van Co. v. Atlas Van Lines, Inc., 792 F.2d 210 (D.C. Cir. 1986) (holding
that antitrust law did not prohibit restraints that eliminate the problem of free riders).
186.
See Raymond T. Nimmer, Standards, Antitrust and Intellectual Property, in Intellectual Property
Antitrust 797 (PLI 1995) (“As a general rule, standards derived and justified by attention to technical, rather than
competitive motivations are not subject to antitrust problems even if they result in a disadvantage to one or several
competitors.”). Of course, it is by no means easy to distinguish legitimate technical concerns from barriers
interposed for anticompetitive reasons. See Transamerica Computer Co. v. IBM Corp., 698 F.2d 1377 (9th Cir.
1983); California Computer Prods., Inc. v. IBM Corp., 613 F.2d 727 (9th Cir. 1979), both of which reject a claim
that IBM has changed its technical interface standards in order to lock out competitors.
++66
CALIFORNIA LAW REVIEW
Finally, some companies may attempt to manipulate the outcome of a standard-setting process
by maneuvering the group into establishing a standard in which the company has a proprietary interest.187 The FTC made this allegation against Dell,188 and others have made it against Unisys.189
Where a party does attempt to capture a standard-setting process in this fashion, the potential anticompetitive consequences seem relatively clear: where network effects exist, controlling a market
standard is substantially more lucrative than in ordinary markets because of the lock-in effects.
Whether this is a problem best treated by antitrust, or instead by a rule of contract law, tort law, or
estoppel, is another matter.190
187.
Capture of standard-setting organizations may also take the form of persuading a standard-setting body
to refuse to allow a competitor’s product to be certified for reasons other than the merits of the product. See Allied
Tube & Conduit Corp. v. Indian Head, Inc., 486 U.S. 492 (1988); American Society of Mechanical Engineers v.
Hydrolevel Corp., 456 U.S. 556 (1982); Radiant Burners, Inc. v. Peoples Gas Light & Coke Co., 364 U.S. 656
(1961). Cf. William J. Curran III, Volunteers . . . Not Profiteers: The Hydrolevel Myth, 33 Cath. U. L. Rev.
147 (1984) (criticizing the Hydrolevel decision).
188.
In re Dell Computer Corp., No. 931-0097 (F.T.C. 1996).
189.
See Lemley, Internet Standardization, supra note 35, at 1087-88 (discussing these allegations); see
also Evans & Reddy, supra note 90, at 17-18; James J. Anton & Dennis A. Yao, Standard-Setting Consortia,
Antitrust, and High-Technology Industries, 64 Antitrust L.J. 247, 250 (1995).
190.
See Lemley, Internet Standardization, supra note 35, at 1090-92 (discussing each possible alternative);
McGowan, supra note 10, at 840-41 (favoring contract alternative). A recent federal circuit case suggests that in the
patent context at least, conduct of this sort may create a legal estoppel barring the patentee from enforcing the patent.
In Wang Laboratories, Inc. v. Mitsubishi Electronics America, Inc., 103 F.3d 1571, 1580-82 (Fed. Cir. 1997), cert.
denied, 118 S.Ct. 69 (1997), the court held that where Wang had sought to establish its patented product in the
market as a standard by encouraging other companies (including Mitsubishi) to produce it, it was estopped from later
suing Mitsubishi for infringement of that patent. It would not be too large a leap to extend this rationale to cases in
[Vol.
YEAR]
SHORT TITLE
67
B. Intellectual Property
“The key formula for the coming age is this: Open, good. Closed, bad.”
Peter Schwartz & Peter Leyden191
The nexus among intellectual property, compatibility, and network effects is quite strong. To
the extent intellectual property rights confer ownership interests in a strong network standard, they
may create durable market power in network markets. Conversely, as we suggested in Part A, the
existence of compatibility between products or standards can in certain circumstances solve problems created by network effects.192 More generally, “the standard” that dominates a network effects
market is not necessarily limited to one firm’s product.
For example, if one believes that
QWERTY typewriter keyboards are dominant network standards, it does not follow that one firm
dominates the typewriter industry, for the simple reason that no one owns the QWERTY standard,
and anyone is free to produce it. Anyone who makes a typewriter keyboard with the same layout
has made a compatible product that will share in whatever network benefits may exist. But if one
firm owned the rights to produce keyboards with a QWERTY layout, it could exclude others from
obtaining the benefits of the standard.
which a patentee encouraged a standards group to adopt and use a standard without disclosing that the standard was
proprietary. See also Stambler v. Diebold, Inc., 11 U.S.P.Q.2d 1709, 1714-15 (E.D.N.Y. 1988) (participant on
standard-setting body who failed to mention patent covering the proposed standard was estopped from later aserting
that patent against members).
191.
Peter Schwartz & Peter Leyden, The Long Boom, Wired, July 1997, at 115, 173.
192.
See supra notes 153-162 and accompanying text.
++68
CALIFORNIA LAW REVIEW
For these reasons, discussions of intellectual property and network effects have tended to focus on arguments that intellectual property rights should be limited in ways that promote compatibility between competing products in network effects markets. We identify and discuss three
variants of this argument below.
1. Arguments for Reverse Engineering
The first, and most limited, argument in favor of compatibility in network industries would
allow reverse engineering in order to achieve compatibility. Reverse engineering is the process of
taking a competitor’s finished product and working backward to determine how it was made.193 A
reverse-engineer can in theory do a variety of things with the information she obtains: duplicate the
product, make a competing product that interoperates with the original product (“horizontal compatibility”), make a complementary product that works with the original product (“vertical compatibility”), or even publish the resulting information for use by others.
The legality of reverse engineering is determined by the relevant intellectual property regime.
Trade secrets law allows reverse engineering of a publicly sold product, regardless of the purpose
to which the resulting information is put.194 Indeed, Supreme Court precedent suggests that the
rule permitting reverse engineering in trade secrets is the economic centerpiece of state trade secret
law and the primary reason that trade secrets coexist peacefully with federal patent and copyright
protection.195 By contrast, patent law provides no protection for those who reverse engineer a pat-
193.
See Uniform Trade Secrets Act § 1, cmt.; Restatement (Third) of Unfair Competition Law § 43, cmt.
b (1995).
194.
See Uniform Trade Secrets Act § 1, cmt.
195.
See Bonito Boats, Inc. v. Thunder Craft Boats, Inc., 489 U.S. 141, 160 (1989) (rejecting a state
statute because it “prohibits the entire public from engaging in a form of reverse engineering of a product in the
[Vol.
YEAR]
SHORT TITLE
69
ented work: if they make, use, or sell the invention, even for commercial research purposes, they
are liable for patent infringement.196 Copyright law occupies a curious middle ground. Until recently, copyright law “permitted” reverse engineering in a very functional sense—one could find
the uncopyrightable ideas hidden in a book without copying the book, and therefore without triggering copyright law at all. As copyrighted works are placed in digital form, this is changing—
public domain. This is clearly one of the rights vested in the federal patent holder, but has never been a part of state
protection under the law of unfair competition or trade secrets.”); Kewanee Oil Co. v. Bicron Corp., 416 U.S. 470,
489-90 (1974) (patent and trade secret law can coexist because “[t]rade secret law provides far weaker protection in
many respects than the patent law. [T]rade secret law does not forbid the discovery of the trade secret by fair and
honest means, e.g., independent creation or reverse engineering. . . .”). Curiously, the traditional rule permitting
reverse engineering is absent from the Economic Espionage Act, the new federal criminal trade secrets statute,
though it does not appear the omission was really considered or intended.
See James H.A. Pooley et al.,
Understanding the Economic Espionage Act of 1996, 5 Tex. Intell. Prop. L.J. 177, 195-97 (1997).
196.
35 U.S.C. § 271(a); see Robert P. Merges et al., Intellectual Property in the New Technological Age
269-70 (1st ed. 1997) (discussing the limited nature of the experimental use defense); Rebecca Eisenberg, Patents and
the Progress of Science: Exclusive Rights and Experimental Use, 56 U. Chi. L. Rev. 1017, 1018-19 (1989) (same).
While in theory the disclosure rquired by 35 U.S.C. § 112 should reduce the need to reverse engineer a patented
product, in practice there are several reasons why a competitor might need to reverse engineer a patented product.
First, she may need access to unpatented compenents of the same device. Second, the patent specification may not
describe the invention in sufficient detail. See Fonar v. General Electric, __ F.3d __ (Fed. Cir. 1997) [check on
cite before going to print] (allowing broad functional descriptions of software). Finally, compatibility may
require an exact picture of the interfaces between two products, and thus may require more detailed information than
that the patentee is forced to disclose.
++70
CALIFORNIA LAW REVIEW
virtually every use of a digital work involves the making of at least one copy.197 In response to this
change, which was first apparent in the context of computer programs, courts have created a limited right to reverse engineer a work (even though a copy is created) in order to extract uncopyrightable information from the work.198 But they have not adopted a universal right to reverse
engineer for any purpose.
197.
See Jessica Litman, The Exclusive Right to Read, 13 Cardozo Arts & Ent. L.J. 29, 34-35 (1994);
accord Jessica Litman, Revising Copyright Law for the Information Age, 75 Or. L. Rev. 19, 21 (1996).
198.
On the legality of reverse engineering in copyright law, compare DSC Communications, Corp. v.
DGI Technologies, Inc., 81 F.3d 597, 601 (5th Cir. 1996); Bateman v. Mnemonics, Inc., 79 F.3d 1532, 1539 n.18
(11th Cir. 1995); Lotus Dev. Corp. v. Borland Int’l, 49 F.3d 807, 817-18 (1st Cir. 1995), aff’d, 516 U.S. 233
(1996), and reh’g denied, 116 S.Ct. 1062 (1996); Atari Games Corp. v. Nintendo of America, Inc., 975 F.2d 832,
843-44 (Fed. Cir. 1992); Sega Enterprises Ltd. v. Accolade, Inc., 977 F.2d 1510, 1527-28 (9th Cir. 1992); Vault
Corp. v. Quaid Software Ltd., 847 F.2d 255, 270 (5th Cir. 1988); Mitel Inc. v. Iqtel Inc., 896 F. Supp. 1050,
1056-57 (D. Colo. 1995), aff’d, 124 F.3d 1366 (10th Cir. 1997) (all endorsing a right to reverse engineer in some
circumstances), with Apple Computer, Inc. v. Franklin Computer, Corp., 714 F.2d 1240, 1253 (3d Cir. 1983) and
Digital Communications Assoc. v. Softklone Distributing Corp., 659 F. Supp. 449 (N.D. Ga. 1987) (older cases
rejecting such a right). The weight of modern authority clearly supports a reverse engineering right.
Most
commentators are in agreement. See, e.g., Band & Katoh, supra note 81; J. Beckwith Burr, Competition Policy and
Intellectual Property in the Information Age, 41 Vill. L. Rev. 193, (1996); Julie Cohen, Reverse Engineering and
the Rise of Electronic Vigilantism: Intellectual Property Implications of “Lock-Out” Technologies 68 S. Cal. L.
Rev. 1091 (1995); Lawrence D. Graham & Richard O. Zerbe Jr., Economically Efficient Treatment of Computer
Software: Reverse Engineering, Protection, and Disclosure, 22 Rutgers Comp. & Tech. L.J. 61 (1996); Dennis S.
Karjala, Copyright Protection of Computer Documents, Reverse Engineering, and Professor Miller, 19 U. Dayton
L. Rev. 975, 1016-18 (1994); Charles R. McManis, Taking TRIPs on the Information Superhighway: International
Intellectual Property Protection and Emerging Computer Technology, 41 Vill. L. Rev. 207 (1996); David A. Rice,
[Vol.
YEAR]
SHORT TITLE
71
In network industries, there is a strong economic argument in favor of permitting reverse engineering in the limited set of cases in which it promotes either vertical or horizontal compatibility
with an industry standard. Because products that can work with an industry standard are included
in the network benefits of widespread adoption, broadly compatible products produced by competitors may allow the market to benefit from a universal standard, while permitting competition
among providers of products incorporating that standard. For this reason, a number of economists
have endorsed the development of a right to reverse engineer, at least in network markets.199 This
is not necessarily a “pure” benefit to competition; rather, it forgoes competition in period one (the
competition that would otherwise have occurred to set the de facto standard) in favor of competition in period two (among competing suppliers of standard-compatible products).200
Of course, in the peculiar economic circumstance that characterizes information markets, more
competition may not always be better. Indeed, the principle behind intellectual property law is that
Sega and Beyond: A Beacon for Fair Use Analysis . . . At Least as Far as It Goes, 19 U. Dayton L. Rev. 1131,
1168 (1994). Other intellectual property statutes expressly include a similar right. See Semiconductor Chip
Protection Act, 17 U.S.C. § 901 et seq.; Brooktree Corp. v. Advanced Micro Devices, Inc., 977 F.2d 1555, 1565-67
(Fed. Cir. 1992).
199.
In addition to the legal commentators cited supra note 198, see, e.g., Jeffrey Church & Roger Ware,
Network Industries, Intellectual Property Rights and Antitrust Policy, in Competition Policy, Intellectual Property
Rights and International Economic Integration (Industry Canada forthcoming 1997); William E. Cohen,
Competition and Foreclosure in the Context of Installed Base and Compatibility Effects, 64 Antitrust L.J. 535, 550
(1996) (“Adoption of a competitive compatibility standard can yield important network effects. It essentially gives
consumers the benefit of other suppliers’ networks.”); Joseph Farrell, Standardization and Intellectual Property, 30
Jurimetrics J. 35 (1989); Katz & Shapiro, Systems Competition, supra note 25, at 111.
200.
See Katz & Shapiro, supra note 25, at 110-11.
++72
CALIFORNIA LAW REVIEW
competition should be sacrificed to some extent in order to give sufficient incentive for innovation.201 Some have argued that intellectual property owners should be given the right to prevent
reverse engineering, because it would further increase their returns.202 However, it does not make
economic sense to give complete control over information to owners of intellectual property. As
Larry Lessig has put it, “[W]hile we protect real property to protect the owner from harm, we protect intellectual property to provide the owner sufficient incentive to produce such property.
‘Sufficient incentive,’ however, is something less than ‘perfect control.’”203 There is ample evidence that the goal of intellectual property law is to balance the incentives given to property owners
against the harm experienced by consumers and next-generation competitors.204 While one might
201.
See generally Mark A. Lemley, The Economics of Improvement in Intellectual Property Law, 75 Tex.
L. Rev. 989, 996-97, 1043 (1997) and sources cited therein; McGowan, supra note 10, at 773-77.
202.
See, e.g., Anthony Clapes, Confessions of an Amicus Curiae: Technophobia, Law and Creativity in
the Digital Arts, 19 U. Dayton L. Rev. 903, 934 (1994); Arthur Miller, Copyright Protection for Computer
Programs, Databases, and Computer-Generated Works: Is Anything New Since CONTU?, 106 Harv. L. Rev. 977,
1029-32 (1993).
203.
Lawrence Lessig, Intellectual Property and Code, 11 St. John’s J. Legal. Comment. 635, 638 (1996).
204.
See, e.g., Graham v. John Deere Co., 383 U.S. 1, 9 (1966) (“The patent monopoly was not designed
to secure to the inventor his natural right in his discoveries. Rather, it was a reward, an inducement, to bring forth
new knowledge.”); Mazer v. Stein, 347 U.S. 201, 219 (1954) (“The economic philosophy behind the clause
empowering . . . patents and copyrights is the conviction that . . . it is the best way to advance public
welfare . . .”); Fogerty v. Fantasy, Inc., 510 U.S. 517, 524 (1994); Feist Publications v. Rural Telephone Serv.
Co., 499 U.S. 340, 349-50 (1991); Stewart v. Abend, 495 U.S. 207, 225 (1990); Bonito Boats, Inc. v. Thunder
Craft Boats, 489 U.S. 141, 146 (1989); Sony Corp. of America v. Universal City Studios, 464 U.S. 417, 429
(1984); Twentieth Century Music Corp. v. Aiken, 422 U.S. 151, 156 (1975); Kewanee Oil Co. v. Bicron Corp.,
[Vol.
YEAR]
SHORT TITLE
73
dispute whether a perfectly formed intellectual property right includes a right to reverse engineer
for any purpose, it seems fundamentally misguided to argue that courts should interpret a law like
copyright—which withholds protection from certain types of information (ideas and facts, for example) precisely in order to make them available to the public—to prevent access to that information.205 The argument is even weaker in the context of strong network effects, where giving such
power to the intellectual property owner converts an open, competitive standard into a proprietary
de facto one, conferring an unexpected windfall on the intellectual property owner206 and reducing
net social welfare.
416 U.S. 470, 476 (1974); Goldstein v. California, 412 U.S. 546, 559 (1973); United States v. Paramount Pictures,
334 U.S. 131, 158 (1948); 17 U.S.C. § 102(b) (idea-expression dichotomy); 17 U.S.C. § 107 (fair use doctrine); 17
U.S.C. § 108 (right to make library copies); 17 U.S.C. § 110 (right to make certain miscellaneous copies and
performances) 17 U.S.C. § 117 (rights to copy computer software); 15 U.S.C. § 1125(c) (protections for news
reporting and noncommercial use in federal dilution statute); 35 U.S.C. § 112 (requirement of public disclosure of
patents); 1 Paul Goldstein, Copyright § 1.14, at 1:40 (1995); L. Ray Patterson & Stanley W. Lindberg, The Nature
of Copyright 120-22 (1991); Cohen, supra note 198, at 1200; Dennis S. Karjala, Federal Preemption of Shrinkwrap
and On-Line Licenses, 22 U. Dayton L. Rev. 511, 512; Mark A. Lemley, Romantic Authorship and the Rhetoric of
Property, 75 Tex. L. Rev. 873, 888-90 (1997); Pierre N. Leval & Lewis Liman, Are Copyrights for Authors or
their Children?, 39 J. Copyrt. Soc’y 1, [recommend pincite] (1991); Jessica Litman, The Public Domain, 39
Emory L.J. 965, 969 (1990); McGowan, supra note 10, at 773-777; Peter S. Menell, An Analysis of the Scope of
Copyright Protection for Application Programs, 41 Stan. L. Rev. 1045, 1082 (1989).
205.
Cf. Robert Kreiss, Accessability and Commercialization in Copyright Theory, 43 UCLA L. Rev. 1, 5
(1995) (suggesting that copyright should be interpreted so as to permit public access to ideas once a work has been
commercialized).
206.
Katz & Shapiro suggest an exception to this: “Since systems competition is prone to tipping, there
are likely to be strong winners and strong losers under incompatibility. Therefore, if a firm is confident it will be
++74
CALIFORNIA LAW REVIEW
The economic argument in favor of reverse engineering to achieve compatibility in network
industries is relatively straightforward in theory. In practice, however, the argument is complicated by the fact that in some industries, particularly the software industry, reverse engineering is
difficult.207 As a result, a right to reverse engineer a computer program will not necessarily prevent
unilateral dominance of a network standard in a software market. The best example of this seems
to be Microsoft’s continued dominance of the operating systems market in software. While the
issue is not entirely free from doubt, and certainly wasn’t seven years ago, it now appears that a
competitor could legally reverse-engineer the current-generation Microsoft operating system, copy
its functional components (at least to the extent they are necessary for compatibility), and build a
competing, “Windows-compatible” operating system. So given the obvious network effects in this
the winner, that firm will tend to oppose compatibility.” Katz & Shapiro, Systems Competition, supra note 25, at
111. While there is certainly historical evidence for this—such as the Bell System’s refusal to interoperate with
telephone competitors at the turn of the century when it had the largest but not yet a dominant market share—it
seems unlikely that any company should be able to make such a prediction before developing an intellectual property
right and therefore unlikely that its research investment decision will be adversely affected by a reverse engineering
rule.
207.
See, e.g, Andrew Johnson-Laird, Software Reverse Engineering in the Real World, 19 U. Dayton L.
Rev. 843, 901 (1994) (reverse-engineering software is technically difficult, time-consuming, expensive, and yields
imperfect results).
[Vol.
YEAR]
SHORT TITLE
75
industry, why hasn’t anyone done so? Or more accurately, given that some have done so,208 why
haven’t they succeeded in competing effectively with Microsoft?209
The answer is complex.210 Several factors have contributed to Microsoft’s continued dominance. First, the legal environment surrounding such a reverse engineering effort has been unclear
until relatively recently. While it is now almost universally accepted that reverse engineering a
computer program in order to obtain access to its unprotectable functional elements and application
program interfaces (APIs) does not violate the copyright laws, virtually none of the cases establishing this principle were decided before 1992.211 Further, Microsoft may have other legal means
at its disposal to prevent such reverse engineering. For example, Microsoft has argued that each of
the 100 million-plus copies of object code it sells are limited distributions of trade secret information subject to a “shrinkwrap license” agreement that prevents reverse engineering, and therefore
that no one can obtain a copy of Microsoft’s operating systems without “agreeing” not to reverse
engineer it.212 More recently, Microsoft has begun to acquire patents governing APIs and similar
208.
Both the IBM OS/2 operating system and the Linux operating system were capable of operating in a
Microsoft Windows-compatible mode. For an entertaining description of the history of Linux, see Glyn Moody,
The Greatest OS that (N)ever Was, Wired, Aug. 1997, at 122.
209.
See generally Rieko Mashima, The Turning Point for Japanese Software Companies: Can They
Compete in the Prepackaged Software Market?, 11 Berkeley Tech. L.J. 429, 431 (1996) (attributing the failure of the
Japanese mass-market software industry at least in part to network effects, which locked in the US first entrants).
210.
We would like to thank all those who lent us their technical expertise in evaluating this market,
including Dale A.J. Dietrich, Mikus Grinbergs, Jessica Litman, and E.L. Oliver.
211.
See supra note 198 (collecting cases on reverse engineering).
212.
For an extreme version of this argument, see Ronald L. Johnson & Allen R. Grogan, Trade Secret
Protection for Mass Distributed Software, in 3 Computer Law Companion 39 (Kyer & Erickson eds., 1995); see
++76
CALIFORNIA LAW REVIEW
interface technology.213 Unlike copyright and trade secret law, patent law has no reverse engineering defense.214 Thus, the legal climate regarding software reverse engineering has never been com-
also Robert Gomulkiewicz & Mary L. Williamson, A Brief Defense of End User License Agreements, 22 Rutgers
Comp. & Tech. L.J. 335 (1996) (a more limited defense of shrinkwrap licenses). The argument that Microsoft’s
object code remains a trade secret despite the distribution of many millions of copies flies in the face of the
fundamental tenets of trade secret law. See Trandes Corp. v. Guy F. Atkinson Co., 996 F.2d 655, 663 n.8 (4th Cir.
1993) (dictum); Young Dental Mfg. Co. v. Q3 Special Prods., Inc., 891 F. Supp. 1345, 1350 (E.D. Mo. 1995)
(characterizing as “completely frivolous” plaintiff’s claim that its publicly sold software was a trade secret);
Religious Technology Center v. Lerma, 908 F. Supp. 1362, 1368 (E.D. Va. 1995) (discussing the availability of
information on the Internet over 10-day period destroyed trade secrecy); Merges et al., supra note 196, at 59-61, 856858; cf. Metallurgical Indus., Inc. v. Fourtek, Inc., 790 F.2d 1195, 1200 (5th Cir. 1986) (considering whether a
limited disclosure to two companies defeated trade secret status); Data General Corp. v. Digital Computer Controls,
Inc., 297 A.2d 433, 436 (Del. Ct. Chanc. 1971), aff’d, 297 A.2d 437 (Del. S. Ct. 1972) (holding that distribution of
approximately 500 copies of program did not destroy trade secrecy). In addition, there is a separate issue as to
whether shrinkwrap licenses are enforceable contracts at all. Compare ProCD, Inc. v. Zeidenberg, 86 F.3d 1447,
1449 (7th Cir. 1996) (enforcing shrinkwrap license) and Maureen O’Rourke, Copyright Preemption After the ProCD
Case: A Market-Based Approach, 12 Berkeley Tech. L.J. 53, 57 (1997) (offering a limited defense of ProCD), with
Novell, Inc. v. Network Trade Center, __ F. Supp. __ (D. Utah 1997) [check on F.Supp. before going to
print] (shrinkwrap licenses unenforceable) and Mark A. Lemley, Intellectual Property and Shrinkwrap Licenses, 68
S. Cal. L. Rev. 1239, 1248-53 (1995) (noting that every other case has refused to enforce shrinkwrap licenses) and
Niva Elkin-Koren, Copyright Policy and the Limits of Freedom of Contract, 12 Berkeley Tech. L.J. 93, 94 (1997)
(criticizing the ProCD decision).
213.
A Lexis search on August 4, 1997 revealed 22 patents assigned to Microsoft which dealt with APIs.
Examples include U.S. Patent No. 5,590,347, for a “Method and system for specifying alternate behavior of a
software system using alternate behavior indicia”; U.S. Patent No. 5,437,006, for a “Spreadsheet command/function
[Vol.
YEAR]
SHORT TITLE
77
pletely certain. This uncertainty can be used to great advantage by aggressive intellectual property
plaintiffs.
Second, reverse engineering an operating system’s program to achieve compatibility in a competing program is particularly difficult. The Microsoft operating systems are complex and interrelated to such an extent that it is not enough merely to build a few hooks into a program. Offering a
competing program with the same functionality would require accurate replication of most, if not
all, of Microsoft’s subroutines.215 Microsoft has strategic incentives to take advantage of this complexity, both by including undocumented calls that may not be easily identifiable, and by playing
on consumer fears of partial incompatibility.
Third, Microsoft regularly changes its operating system. It has upgraded its operating system
software on a regular basis over the last 15 years, and on several occasions has introduced an entire new generation of operating system, despite the difficulty of migrating the installed base of users from an old program to a new (albeit compatible) one. A competitor would have to reverse
engineer each new program and alter its own program to maintain compatibility with the new generation Microsoft product. Ideally, it would have to do it quickly, since even a six-month time lag
in introducing a competing product can be fatal in a market which moves as fast as software. Regardless whether Microsoft deliberately introduced changes in its operating systems to deter competitors (such conduct may also be consistent with desirable product innovation), the evolution of
its programs had that effect.
capability from a dynamic-link library”; and U.S. Patent No. 5,430,878, for a “Method for revising a program to
obtain compatibility with a computer configuration.”
214.
See supra note 196 and accompanying text.
215.
See Harz, supra note 175, at 211-13 (describing this problem, and invoking Clarke’s Law to suggest
that a sufficiently hard problem of this sort might as well be impossible).
++78
CALIFORNIA LAW REVIEW
Finally, Microsoft simply doesn’t charge very much for its operating system.
The price
(about $99 for an off-the-shelf copy) is unquestionably above marginal cost, as it is with all software protected by an intellectual property right, but it doesn’t seem to reflect the full power of the
network effects Microsoft has harnessed. A competitor simply may not find it worthwhile to expend the effort required to build and maintain a compatible operating system, facing possible legal
action, in order to sell its product at a relatively low price. This is especially true given Microsoft’s
history of competing for market share by lowering price (or even giving products away for free)
when challenged.216
That reverse engineering is not a perfect route to interoperability may be either a good or a bad
thing, depending on one’s perspective. Those who worry that rapid imitation will leave little incentive for intellectual property owners to create new works may take some solace in the fact that
reverse-engineers will have to spend time and money discovering the ideas contained in a new
piece of software, giving some latitude to earn supracompetitive returns before the ideas are discovered.217 Network effects may actually enhance this technological first-mover advantage, because the tipping effect may produce a rapid supracompetitve return before imitators can effectively
reverse engineer.218 Further, reverse engineering does not permit competition within a standard in
216.
See Peter H. Lewis, Microsoft to Give Away Web Server Software, N.Y. Times, Feb. 6, 1996, at C6
(describing Microsoft’s pricing policy regarding Web browsers).
217.
See, e.g., Pamela Samuelson et al., A Manifesto Concerning the Legal Protection of Computer
Programs, 94 Colum. L. Rev. 2308, 2430 (1994) (arguing that if reverse engineering were cheap and easy,
insufficient incentive might be provided to software developers; making a parallel argument for design know-how
that is readily apparent from a casual inspection of the program).
218.
advantage”).
See Cohen, supra note 199, at 541 (“network effects can be significant sources of first-mover
[Vol.
YEAR]
SHORT TITLE
79
all cases, which may add some further incentive to create new products, or at least to try to win a
standards competition.219
The imperfection of reverse engineering as a device for opening standards to competition has
negative consequences for consumers. If effective competition for a Microsoft-compatible operating system existed, the price of the operating system presumably would fall, and more consumers
would adopt the standard, to the benefit of all. Nonetheless, it appears that the threat of entry
drives Microsoft to develop new product upgrades and to price its product at a fairly reasonable
level. Perhaps even where the legality of reverse engineering does not lead to effective competition
within a standard, it nonetheless operates as a constraint on the power that standard-setters would
otherwise exercise in network industries.
2. Arguments Against Protecting Interface Components
In part because reverse engineering is not a perfect solution to the problem of standards dominance, a number of commentators have argued that intellectual property should not protect interface
components at all, allowing unlimited access to such components by competitors. 220 More gener-
219.
Indeed, some companies have gone to the extreme of making their own product incompatible with
itself, precisely in order to make copying less lucrative. See, e.g., David Lazarus, DVD’s World (And Welcome to
It), Wired, July 1997, at 42 (noting that manufacturers of digital versatile discs have made different, incompatible
disks for marketing in different parts of the world). From a social welfare perspective, this is disastrous in a network
market. That the companies involved are willing to forego significant internal network benefits suggests precisely
how concerned they are about easy imitation of their products.
220.
See, e.g., Farrell, supra note 154, at 369-70; Peter S. Menell, An Analysis of the Scope of Copyright
Protection for Application Programs, 41 Stan. L. Rev. 1045, 1066-72 (1989); Peter S. Menell, The Challenges of
Reforming Intellectual Property Protection for Computer Software, 94 Colum. L. Rev. 2644, 2652-53 (1994).
++80
CALIFORNIA LAW REVIEW
ally, to the extent the existence or scope of an intellectual property right in a standard is undetermined, courts have considered network effects in deciding whether or not to grant a new or
stronger form of intellectual property protection to the standard-setter.221
The bellwether case for software interface components was Lotus Dev. Corp. v. Borland
Int’l. 222 Stripped to its essence, the case involved a copyright dispute between first-and secondgeneration spreadsheet vendors over the ability of the first-generation developer to control the program and user interfaces of the spreadsheet. The very basic, pre-Windows Lotus 1-2-3 interface
involved two lines of word commands that appeared on each screen at any given time, and which
were part of an overall menu structure or hierarchy. The total number of commands in the menu
command hierarchy was 469; all the commands were intuitive one-word instructions like “Graph”
and “Print.” Borland’s Quattro Pro spreadsheet, by contrast, was a more sophisticated, Windowsbased program with a much more detailed user interface. Lotus sued Borland because the Quattro
Pro spreadsheet included within it “Key Reader” features that allowed users to run their own 1-23-based macros in Quattro Pro, and an “Emulation Mode” that allowed users to use 1-2-3 commands to operate the Quattro Pro spreadsheet.
221.
The clearest example is United States Golf Ass’n v. St. Andrews Systems, 749 F.2d 1028, 1037-41
(3d Cir. 1984), where the court rejected the USGA’s claim to own a property right in its standard handicapping
system. Other cases exist in the computer industry, however; see infra notes 222-242 and accompanying text..
222.
49 F.3d 807 (1st Cir. 1995), aff’d by equally divided Court, 116 S. Ct. 804 (1996). The case has
been discussed in a great deal of factual detail elsewhere; we will not repeat this background here. See, e.g., Glynn
S. Lunney, Jr., Lotus v. Borland: Copyright and Computer Programs, 70 Tulane L. Rev. 2397 (1996); David R.
Owen, Interfaces and Interoperability in Lotus v. Borland: A Market-Oriented Approach to the Fair Use Doctrine, 64
Fordham L. Rev. 2381 (1996).
[Vol.
YEAR]
SHORT TITLE
81
The district court found that the overall menu command hierarchy (and at least some of its
constituent elements) was copyrightable, and that Borland necessarily copied that hierarchy in both
the Emulation and Key Reader features.223 The First Circuit reversed, with the majority ruling
broadly that Lotus’ menu command hierarchy was not copyrightable at all because it was a
“method of operation” of the 1-2-3 program.224 Judge Boudin, concurring, offered an alternative
rationale, one at least implicitly based in network effects. In his view, the key to the case was that
Borland was not seeking to trade on Lotus’ contributions,225 but merely to assist the user in migrating her data from 1-2-3 to Quattro Pro, and to prevent her from being “locked in” by her investment in learning the 1-2-3 program.226
Argument in the Lotus case focused a great deal of attention on the network effects of Lotus 12-3’s original dominant market position. A prominent group of economics professors took the unusual step of filing an amicus brief with the Supreme Court in the case.227 They argued that computer program interfaces exhibited strong network effects:
223.
The district court’s decisions are reported at Lotus Dev. Corp. v. Borland Int’l, Inc., 788 F. Supp. 78
(D. Mass. 1992); 799 F. Supp. 203 (D. Mass. 1992); 831 F. Supp. 202 (D. Mass. 1993); and 831 F. Supp. 223
(D. Mass. 1993).
224.
49 F.3d at 819.
225.
Id. at 821 (Boudin, J., concurring) (“Borland’s use is privileged because . . . it is not seeking to
appropriate the advances made by Lotus’ menu.”).
226.
Id. See also Hamilton & Sabety, supra note 34, at 274-75. Hamilton & Sabety endorse this result,
noting that giving Lotus the “capacity to block a migration path for software consumers is anticompetitive rather
than the result of a legitimate copyright monopoly.” Id. at 275.
227.
See Lotus Dev. Corp. v. Borland, Int’l, Inc., Brief Amicus Curiae of Economics Professors and
Scholars in Support of Respondent, No. 94-2003.
++82
CALIFORNIA LAW REVIEW
Network effects are important in software markets. Users want to share data files
and programs such as macros; they want to work on machines owned by others;
they want access to a wide selection of complementary products (including thirdparty manuals, consulting services, training courses, and add-on software). Certain aspects of programs must be identical in order for users of different programs
to share these network benefits; these aspects will predictably include “interfaces”
and aspects of a program that define a language, such as a macro language.228
These problems could be solved, in the economists’ opinion, by allowing competitors to use the
program interfaces.
Whether or not network effects and switching costs limit competition depends crucially on whether or not vendors have proprietary control of the interfaces. If interfaces are public, competitors can make their products compatible, and users will be
able to choose a program on the basis of its quality and price rather than on switching costs and installed base of users. If interfaces are protected by copyright, the
copyright holder can prevent competitors from making their products compatible.
In this way the intellectual property treatment of interfaces crucially affects the nature of competition.229
228.
Id. at 5.
229.
Id. at 8.
[Vol.
YEAR]
SHORT TITLE
83
Others have made similar arguments against protecting program interfaces, both in briefs and in
academic commentary.230
This argument is attractive for the same reasons as arguments in favor of allowing reverse engineering: both would permit competitors to build different but interoperable programs, effectively
opening de facto standards to competition in the same way that an open standard-setting group
permits competition in products within the standard. Competition means that price declines toward
marginal cost, increasing efficiency and inducing more people to join the network standard to the
benefit of all. 231 If this is the goal, though, we ought not limit ourselves to copyright. Program
interfaces might also be locked up by being patented, or by being kept secret from competitors, as
seems to have been the case with Microsoft’s operating system. If we were truly to permit competition within de facto standards, we would have to deny all forms of intellectual property protection
to the interfaces that allow access to such standards.232
230.
See, e.g., Mitel v. Iqtel, 896 F. Supp. 1050, 1056 (D. Colo. 1995), aff’d, 124 F.3d 1366 (10th Cir.
1997); Dennis S. Karjala & Peter S. Menell, Brief Amicus Curiae: Applying Fundamental Copyright Principles to
Lotus Development Corp. v. Borland International, Inc., 10 High Tech. L.J. 177, 178 (1995); Farrell, supra note
154, at 9-10; Hamilton & Sabety, supra note 34, at 274-77. Cf. Stephen M. McJohn, Fair Use of Copyrighted
Software, 28 Rutgers L.J. 593, 634-35 (1997) (arguing for a stronger fair use defense in software cases). But see
David Friedman, Standards as Intellectual Property: An Economic Approach, 19 U. Dayton L. Rev. 1109, 1110-11
(1994) (suggesting that proprietary control over market standards should not be considered problematic).
231.
See Farrell, supra note 154, at 3.
232.
See Robert J. Levinson, Concerns Raised by Recent Software-Related Antitrust Cases, 41 Antitrust
Bull. 43, 48 (1996) (suggesting a regime in which intellectual property law does not prohibit competitors from
creating software compatible to existing standards); Julie Cohen, supra note 198, at 1168-81 (arguing against broad
++84
CALIFORNIA LAW REVIEW
Yet interfaces themselves may exhibit the market failures that justify intellectual property protection. If interfaces are excluded from intellectual property protection entirely, it is possible that
companies will not invest time or effort to develop them.233 This is particularly likely if the interfaces at issue are complex and difficult to design, as is often the case with computer software.234
patentability of program interface standards and “lock-out” devices); Lemley & O’Brien, supra note 38, at 301-03
(arguing that patent law should be stricter toward attempts to patent simple interface connections).
Alternatively, one could endorse a system of compulsory licensing at a government-set rate, allowing
some compensation to the owners of the standard but preventing them from denying access to the standard. See
Peter S. Menell, Intellectual Property: General Theories, in The New Palsgrave Encyclopedia of Law and
Economics (forthcoming 1998) (“compulsory licensing may be justified in particular circumstances to enable the full
realization of network externalities”). Compulsory licensing has its own set of problems, however, which are treated
elsewhere.
233.
See Friedman, supra note 153, at 1121; accord Kenneth Dam, Some Economic Considerations in the
Intellectual Property Protection of Software, 24 J. Legal Stud. 321, 338 (1995). Friedman is probably wrong to
suggest that investment in making an interface into a standard should be rewarded, however. See Friedman, supra, at
1121-22. Not only will effort invested in market share be its own reward if the standard-setter has proprietary control
over the standard, but experience in a number of industries has shown that investment by a single firm in
determining a standard may be unnecessary, since if the putative standard is opened to competitors it may actually do
better than other, competing standards that are closed. The success of IBM-compatible PCs and VHS VCRs are
well-known examples.
234.
For example, the authors of the Manifesto worry that program design information will be
underproduced because current copyright law gives insufficient protection to such information. See Samuelson et
al., supra note 217, at 2398-99. See also Baseman et al., supra note 74, at 270 (IBM spent $2 billion developng
OS/2). But see Leeds, supra note 160, at 656 (suggesting that consumer brand identification may mean that first-
[Vol.
YEAR]
SHORT TITLE
85
An example of this difficulty is the First Circuit’s result in the Lotus case itself. While it may make
good economic sense to allow a transformative improver like Borland to build on the Lotus 1-2-3
template, thus allowing existing 1-2-3 users to migrate without incurring significant switching
costs, the court’s opinion would also exempt from liability a “clone”-maker like Paperback Software, which did nothing but imitate the Lotus user interface and menu command hierarchy.235 The
equities of the two cases seem very different, largely because Borland invested in improving rather
than merely copying the Lotus interface. 236 One might conceive of Paperback’s clone copy as a
beneficial effect of opening the standard to competition. The problem is that, at least in this case,
virtually all of the value Lotus added is contained in the interface the court declares unprotectable.
One might rebut the traditional case for intellectual property in network markets, however, if
one can demonstrate that the network effects themselves will ensure an adequate return to the initial
creator even absent intellectual property protection. The argument would be that the creator of a
program component in a network market will benefit from the additional adoptions caused by unauthorized copying, because it increases the chance the component will become a standard, and
increases the value to existing consumers of using the creator’s product. Some economists have
indeed made this argument.237 The argument seems best suited to circumstances where copying is
movers still retain advantages, and therefore that there may still be sufficient incentives to invest in developing new
standards).
235.
See Lotus Dev. Corp. v. Paperback Software, 740 F. Supp. 37 (D. Mass. 1990).
236.
See Lemley, supra note 201, at 1079-81; Mark A. Lemley, Convergence in the Law of Software
Copyright?, 10 High Tech. L.J. 1, 31-32 (1995); McGowan, supra note 38, at 848 n.310; Owen, supra note 222 at
2418-23 (all making this point).
237
See Lisa N. Takeyama, The Welfare Implications of Unauthorized Reproduction of Intellectual
Property in the Presence of Demand Network Externalities, 42 J. Indus. Econ. 155 (1994).
++86
CALIFORNIA LAW REVIEW
imperfect, or at least costly, such as reverse engineering. If copying were cheap and easy, and if
there were no first mover advantages,238 it is hard to see how the original creator could capture any
benefit from the adoption of her product as a standard.
A more general problem with eliminating intellectual property protection for interfaces is that it
is far from clear that all or even most computer programs are standards in network effects markets.
The operating system market for personal computers seems characterized by network effects, as
discussed above.239 Peter Menell has also argued that user interfaces exhibit network effects, since
people want to use a single, standardized interface with which they are already familiar;240 at least
one court has endorsed his view.241 The nature of the network effect is different in the two markets, however: operating systems exhibit network effects because application programmers need
to write compatible software, while user interfaces exhibit only the “learning effect” of saving users from having to learn how to operate multiple systems. And there may well be program interface components that do not exhibit network effects at all.
A rule precluding protection for
program interfaces in every case cannot be justified by network effects if those effects are not
themselves present in every case. A more nuanced approach, like the one suggested above that
238
For a discussion of first mover advantages in the software context, see Lemley & O’Brien, supra note
38, at 274-75.
239.
See supra notes 72-83 and accompanying text. See also Peter S. Menell, Tailoring Legal Protection
for Computer Software, 39 Stan. L. Rev. 1329, 1357-58 (1987).
240.
Peter S. Menell, An Analysis of the Scope of Copyright Protection for Application Programs, 41
Stan. L. Rev. 1045, 1066-71 (1989).
241.
See Apple Computer, Inc. v. Microsoft Corp., 799 F. Supp. 1006, 1025 (N.D. Cal. 1992), aff’d, 35
F.3d 1435 (9th Cir. 1994).
[Vol.
YEAR]
SHORT TITLE
87
distinguishes the Borland and Paperback cases, seems warranted. Indeed, Menell himself has acknowledged as much in his later work.242
Evidence of such a nuanced approach can be found in a non-software case involving the proposed extension of intellectual property law to protect a standardized golf handicapping system. In
United States Golf Association v. St. Andrews Systems, 243 the USGA, the “governing body of
amateur golf in the United States,”244 sought intellectual property protection for its formula for
equalizing competition among golfers of different skill levels by assigning a “handicap” score to
each. The defendant, Data-Max, developed a computer program that calculated a golfer’s USGA
handicap. The court disposed readily of USGA’s trademark infringement claim, finding that the
formula was functional and hence not itself capable of trademark protection.245 The most interesting part of the opinion was a thoughtful discussion of USGA’s argument that the common-law tort
of misappropriation ought to be extended to cover Data-Max’s conduct.246 The court refused to
extend the doctrine beyond the classic case of direct competition. It reasoned:
242.
See Menell, supra note 220, at 2647-48 (“Intellectual property protection plays a critical role in
determining the size and quality of these networks. Software products or attributes affecting the size of networks
should, therefore, not generally receive intellectual property protection unless they require significant research
efforts.”) (emphasis added). Of course, determining ex ante which inventions meet this test has not proven to be an
easy task.
243.
749 F.2d 1028 (3d Cir. 1984).
244.
Id. at 1030.
245.
Id. at 1034.
246
Whether state-law misappropriation doctrine could be extended in the face of federal preemption is a
different question than the one the court addressed. On this issue, see NBA v. Motorola, 105 F.3d 841 (2d Cir.
1997) (only a limited misappropriation doctrine survives preemption).
++88
CALIFORNIA LAW REVIEW
The public acceptance of the U.S.G.A.'s handicap formula stems from the golfing
public's desire to have a uniform system of quantifying recent performances in a
way that will allow equitable competition among golfers of differing abilities. The
U.S.G.A., in furtherance of its role as the governing body of amateur golf, has
provided such a system and, in the absence of a better system, the public has apparently accepted it. Under this state of affairs, the emergence of a single standard becomes largely a function of the need for uniformity. To require Data-Max to use a
different formula would effectively destroy its ability to provide a handicapping
service, since the U.S.G.A. formula is widely accepted by the golfing public. The
purpose of a handicap is comparison between golfers, and handicaps based on different formulas cannot be readily compared.
Because the U.S.G.A. formula is the equivalent of an "industry standard"
for the golfing public, preventing other handicap providers from using it would effectively give the U.S.G.A. a national monopoly on the golf handicapping business. Where such a monopoly is unnecessary to protect the basic incentive for the
production of the idea or information involved, we do not believe that the creator's
interest in its idea or information justifies such an extensive restraint on competition. This case provides a good example of why such a restraint would harm the
golfing public. Data-Max has expended time and creative energy in devising its
own products and services. It has not only created the program used to calculate
handicaps by computer, but has devised a handicapping service which improves on
that provided by the U.S.G.A., at least to the extent that Data-Max provides a
golfer with a fresh handicap faster than the U.S.G.A. does.
In addition, the
[Vol.
YEAR]
SHORT TITLE
89
U.S.G.A. has not been completely deprived of the opportunity to be compensated
for its "good will" in connection with the handicap formula. To the extent that the
approval of the U.S.G.A. would enhance the value of "instant handicaps," the
U.S.G.A. has an opportunity, if it wishes to exercise it, of offering either DataMax or rival the use of the U.S.G.A. name in marketing its products and services.247
The Court bases its argument in this case on network effects, but limits the discussion to a specific
factual situation similar in certain respects to the Lotus case. It is also similar to Lotus in another
respect; in both cases, the court was asked to rule on a novel issue of intellectual property law. It
is much more difficult to find a case considering network effects arguments as a reason to depart
from or modify established intellectual property law.
3. Compatibility as an Antitrust Issue
An alternative approach to these issues involves the application of antitrust (or analogously,
misuse) law to computer markets that exhibit network effects. Not surprisingly, in light of what
we have seen in antitrust law,248 network effects arguments in intellectual property have been most
successful when presented in an antitrust context. We have already seen one specialized example—the Dell Computer case, where the FTC blocked what it viewed as the strategic assertion of
intellectual property rights to control a standard-setting organization.249 But there are other exam-
247.
Id. at 1040-41 (footnotes omitted).
248.
See supra notes 69-190 and accompanying text (noting the widespread use of network effects
arguments in antitrust cases).
249.
In re Dell Computer Corp., No. 931-0097 (F.T.C. 1996).
++90
CALIFORNIA LAW REVIEW
ples of cases where the government used its antitrust power to compel interoperability or the licensing of intellectual property. 250 Virtually none of these involve “compulsory licensing” of intellectual property in a strict sense; rather they involve government-imposed conditions on a merger
of two companies.251
While the government has argued few of these cases on network effects grounds, the license
imposed by consent decree in In re Silicon Graphics, Inc.252 is expressly based on the Commission’s concern that once Silicon Graphics had acquired software makers Alias and Wavefront,
“entry [into the market for workstations running entertainment graphics software] would be unlikely. Marketing a technically comparable or even an improved combination of non-SGI workstations with entertainment graphics software, other than that of Alias or Wavefront, would be
difficult, time consuming, and not likely to occur because of the extensive installed user base of
SGI workstations with Alias, Wavefront and SoftImage entertainment graphics software.” 253 The
Commission’s solution to this network effect is to compel the licensing of the Alias and Wavefront
software on nondiscriminatory terms.254 In dissent, Commissioner Starek suggests that requiring
nondiscriminatory licensing is problematic, pointing to the cross-subsidization problems encoun-
250.
See, e.g., United States v. Thomson Corp., 949 F. Supp. 907 (D.D.C. 1996) (conditioning
Thomson’s purchase of West Publishing Corp. on West’s agreement to license its (allegedly copyrighted) pagination
system to competitors); James P. Love, A Free Trade Area for the Americas: A Consumer Perspective on Proposals
as they Relate to Rules Regarding Intellectual Property, http://www.cptech.org/pharm/belopaper.html#examples
(visited June 10, 1997) (collecting cases).
251
Id.
252.
No. C-3626 (F.T.C. Nov. 14, 1995).
253.
Id. at 4.
[Vol.
YEAR]
SHORT TITLE
91
tered in United States v. AT&T. 255 While Commissioner Starek is certainly correct that requiring
licensing of intellectual property on nondiscriminatory terms is not problem-free, and may require
continued regulatory oversight, it does not follow that it is never an appropriate antitrust remedy.
In particular, the majority in Silicon Graphics reasoned that allowing the vertical combination of
SGI with the two software firms (and hence the horizontal combination of Alias and Wavefront
themselves) produced substantial market efficiencies. The FTC’s remedy arguably preserved those
efficiencies while opening the standard to effective competition.256
Even where the government does not bring an action against merging firms or standard-setting
groups, the existence of open license policies may play a role in avoiding antitrust scrutiny. For
example, the Department of Justice recently cleared a potentially troubling pool of patents relating
to video data compression, even though the patents would be licensed only as a group. Joel Klein,
the head of the antitrust division, noted that “the proposal does not appear to have an anticompeti-
254
Id.
255.
Id. at 4 & n.9 (Starek, Comm., dissenting).
256.
By contrast, Commissioner Azcuenaga in her separate dissent would have blocked the merger of Alias
and Wavefront with Silicon Graphics on the grounds that combining Alias and Wavefront increased the horizontal
market power of the combined firm. Id. at 1 (Azcuenaga, Comm., dissenting). This alternative remedy would have
dealt with concerns about the market power produced by a combination, but would not have captured whatever
network efficiencies were attributable to the proposed merger. A virtually identical case—with a similar result, and
the same split among the commissioners—is In re Cadence Design Systems. No. 971-0033 (F.T.C. May 6, 1997).
++92
CALIFORNIA LAW REVIEW
tive effect on rivals because the license will be available to all applicants on the same terms and
conditions.”257
Another approach to the same issues can be found in the patent and copyright misuse doctrines.
In part because of their flexibility—the copyright misuse doctrine has no statutory
bounds,258 and the patent misuse doctrine has only a set of specific limitations rather than a general
statutory scope 259—the misuse doctrines may be well-suited to limiting the reach of intellectual
property law in network cases. Indeed, Julie Cohen has argued for a misuse-based limit on both
patented and copyrighted “lockout” devices that is distinct from the antitrust principles normally
applied in misuse cases.260 The Fifth Circuit adopted such an approach in DSC Communications v.
DGI Technologies. 261 In that case, the court held that it would likely be copyright misuse for DSC
to assert that any testing of microprocessor cards by a competitor infringes its copyright on the
257.
Division Clears Patent Pooling, Licensing for Compressed Video Data, Antitrust & Trade Reg. Rpt.,
July 10, 1997, at 31. See also Leeds, supra note 160, at 641-42 (describing antitrust scrutiny of the DVD
consortium, with the same result).
258.
See Lasercomb America v. Reynolds, 911 F.2d 970, 973 (4th Cir. 1990) (“[A] misuse of copyright
defense is inherent in the law of copyright just as a misuse of patent defense is inherent in patent law.”).
259.
See 35 U.S.C. § 271(d); see generally Mark A. Lemley, Comment, The Economic Irrationality of the
Patent Misuse Doctrine, 78 Calif. L. Rev. 1599, 1610 (1990) (commenting on the odd codification of the patent
misuse doctrine).
260.
Cohen, supra note 198, at 1190-98. But see Marshall Leaffer, Engineering Competitive Policy and
Copyright Misuse, 19 U. Dayton L. Rev. 1087, 1104-06 (1994) (arguing that copyright misuse should not protect
efforts to reverse engineer). Cf. Robert P. Merges, Patent Law and Policy 1183-87 (2d ed. 1997) (noting possible
scope for the patent misuse doctrine outside the bounds of antitrust law).
261.
81 F.3d 597, 601 (5th Cir. 1996).
[Vol.
YEAR]
SHORT TITLE
93
microcode contained in those cards.262 The court found that DSC was “attempting to use its copyright to obtain a patent-like monopoly over unpatented microprocessor cards” by locking out competitors from producing and testing compatible cards. There is no indication in DSC that the
decision is based on network effects, but it seems clear that the court was willing to apply misuse
doctrine to protect reverse engineering in order to achieve system compatibility. It would certainly
be possible to tailor this approach, finding misuse only where an intellectual property owner attempted to use the law to lock up access to a network standard.263
One can agree or disagree with the Commission’s assumption in Silicon Graphics that the
market for entertainment graphics software is characterized by strong network effects, so that
locked-in consumers will be unlikely to switch to competing software. Certainly the Commission’s opinion does not provide much detailed economic evidence in support of this conclusion.
One can similarly disagree about the wisdom of applying misuse doctrine in any given case. But
the critical point about the Silicon Graphics and DSC cases is that approaching these issues using
262.
The testing was argued to be copyright infringement because it made a temporary “copy” of the
copyrighted code in RAM memory. See MAI Sys. Corp. v. Peak Computing, Inc., 991 F.2d 511, 518 (9th Cir.
1993). The DSC decision suggests that the Fifth Circuit would not agree with the MAI court’s interpretation of the
copyright laws.
263.
As some commentators have noted, the remedy imposed in misuse cases—refusing to enforce the
intellectual property right—makes no economic sense. See Lemley, supra note 259, at 1614-20; Church & Ware,
supra note 199, at 34-35. But the limited, quasi-equitable nature of the remedy might actually be advantageous in
network cases. If an intellectual property right is unenforceable only in the circumstances in which it is used to
capture control over a network standard, access to the standard might be preserved without either punishing the
intellectual property owner under antitrust law or depriving it of all use of its intellectual property. Cf. Julie Cohen,
supra note 198, at 1193 (suggesting modification of the patent misuse remedy along similar lines).
++94
CALIFORNIA LAW REVIEW
the detailed factual economic analysis that should characterize antitrust rule of reason inquires is
preferable to resting on the more general principle that the intellectual property laws contemplate the
possibility of supracompetitive returns.264 There is no general requirement in antitrust law that the
owner of an intellectual property right make it available to potential competitors, even when the intellectual property owner is in a position of market power.265 One can disagree about whether such
an obligation is ever appropriate in a specific case.266 But rule of reason methodology—essentially
the approach embodied in Judge Boudin’s concurrence, though under a different rubric—avoids
the problem the First Circuit confronted in the Lotus case: how to craft an intellectual property rule
that will enhance social welfare in network effects cases without inflicting too much damage on the
fundamental tenets of intellectual property law in the majority of cases where network effects do
not play a role.
264.
See Data General Corp. v. Grumman Systems Support, 36 F.3d 1147, 1187 (1st Cir. 1994)
(establishing a rebuttable presumption against finding an antitrust violation on the basis of a refusal to license
intellectual property, but noting that “there may be rare cases in which imposing antitrust liability is unlikely to
frustrate the objectives of the Copyright Act”). Band and Katoh suggest that de facto standards in network effects
markets ought to be considered one of these “rare cases.” Band & Katoh, supra note 81, at 47.
265.
See, e.g., Data General Corp. v. Grumman Systems Support, 36 F.3d 1147, 1188 (1st Cir. 1994);
Berkey Photo, Inc. v. Eastman Kodak Co., 603 F.2d 263, 283 (2d Cir. 1979). But see Leeds, supra note 160, at
658 (suggesting that the reasoning of Berkey Photo “breaks down in the presence of network externalities”).
266.
Indeed, we have. Compare McGowan, supra note 10, at 850 (“the essential facilities doctrine has no
place in the legal regime being crafted to regulate software”, at least given the presence of alternatives), with Lemley,
Internet Standardization, supra note 35, at 1084-86 (arguing that while essential facilites doctrine should be rarely
used, it may be appropriate in some cases).
[Vol.
YEAR]
SHORT TITLE
95
4. Conclusions
Several things are worth noting about the role network effects have played in intellectual property arguments. First, unlike the antitrust cases, network effects arguments in intellectual property
cases are virtually all on one side of the dispute—the defendant’s. To date, no one has made an
argument for expanded intellectual property protection based on network effects. 267 Second, network economics has not penetrated as far into legal doctrine in intellectual property as it has in antitrust law.
Though a number of cases have adopted legal rules based explicitly on their
consideration of network effects, these all exist at the margins of intellectual property law—where
an argument is made for expanding intellectual property rights beyond their existing boundaries, or
at least where the scope of those rights has not yet been determined. Even the one case that might
be said to cut back on well-established legal protection (Lotus v. Borland) involved what the court
perceived to be a novel, unresolved legal issue. That network effects arguments are most successful only at the margins of intellectual property law may suggest that courts hearing intellectual
property cases have not integrated economic thought into their decisionmaking process as thoroughly as in antitrust law. That a major use of network effects arguments in intellectual property
cases has been in the antitrust and misuse contexts further bolsters this view.
C. Government Standard-Setting268
We have so far discussed industries in which standards either emerge from competition to set
them, or are set or encouraged by private organizations that exist for that purpose. An alternate
267.
A few commentators have suggested that strong property rights were still appropriate in such markets,
however. See Dam, supra note 233, at 372; Friedman, supra note 153, at 1117.
268.
1062-64.
This section elaborates issues first introduced in Lemley, Internet Standardization, supra note 35, at
++96
CALIFORNIA LAW REVIEW
possibility is that the government might identify and set the appropriate standards and compel all
participants in the market to comply. The government does this from time to time.269 For example,
the Federal Communications Commission sets standards for interconnection between telephone
networks and standards governing the use of products that might interfere with broadcast communications.270 More recently, the United States government stepped into the debate over the proper
standard for high definition television (HDTV), selecting a standard that unified U.S. development
work but was at odds with other standards adopted in Japan and Europe.271 And government agencies such as the Advanced Research Projects Agency and the National Science Foundation played a
role in the development of the Internet, including the creation of Internet interconnection protocols.
Indeed, private Internet standard-setting groups such as InterNIC and the IETF were once government-sponsored standards organizations.272 Government-set standards have some of the same
appeal as those set by private organizations: a wasteful competition to set a de facto standard is
avoided, and the government can presumably mandate open access to the technical interface, permitting competition within the standard.
Whether government standard-setting is a good idea or a bad idea depends on the characteristics of the market. In some industries, it is critical that a single standard be set immediately, almost
269.
See Andrew Updegrove, Consortia and the Role of the Government in Standard Setting, in Standards
for Information Infrastructure, supra note 154, at 321.
270.
F.C.C. Rules, 47 C.F.R. § 68.1.
271.
See Denise Caruso, Debate Over Advanced TV Gives the F.C.C. a Chance to be Assertive, N.Y.
Times, June 17, 1996, at C5; F.C.C. Proposes Standards for Digital Television, N.Y. Times, May 10, 1996, at C4.
272.
Government may also play a standard-setting role as a market participant, since it is often one of the
largest customers of a given product. In most instances, however, this public procurement does not create the same
problems as government-mandated standards.
[Vol.
YEAR]
SHORT TITLE
97
without regard to what that standard is. Where the need for a high degree of uniformity in adoption of a standard coincides with the technical simplicity of the standard, government action may
sometimes be desirable. Consider roads. Whether all traffic should drive on the left or the right is
essentially an arbitrary decision. It is far more important that everyone obey the same rule than that
the “best” side be chosen. While it is certainly possible to let the market solve this problem by
means of a standards competition—where cars optimized for either left-or right-side driving win
out through the tipping effect—the social cost of holding such a competition would be quite large.
A government-set standard is preferable in this instance because the precise standard chosen does
not matter very much, and the government can impose various legal sanctions (including criminal
ones) to enforce uniform compliance with the standard.
Of course, not all networks share these characteristics. There are several reasons why government control of the standard-setting process should not be encouraged in the ordinary case,
where the choice of standard makes a difference for social welfare. First, government agencies are
generally composed of career public servants, not market participants. As a result, these agencies
may not involve the most qualified individuals in the standard-setting process.273 This is an inherent danger of bureaucracy, particularly when it attempts to regulate such a fast-moving area of
commerce as the Internet. Government standard-setting groups may be slow and may not always
have access to the best information. Thus, even with the best of intentions, a government standard-setting organization may simply pick what is objectively a poor standard, as almost happened
in the case of the United States HDTV standard around 1990. Only by an accident of timing did
273.
See McGowan & Lemley, supra note 158, at 336-37.
++98
CALIFORNIA LAW REVIEW
the government adopt a digital HDTV standard, rather than an analog standard that would have
been immediately obsolete.274 Even then, the issue remained unresolved in 1997.275
Second, government-set standards may prove durable even when they are demonstrably illconceived.276 While the market eventually will replace an inefficient standard by "leapfrogging,"
there is no guarantee that the government will do the same.277 And since in our economy the market
is the chief determinant of what is in fact efficient, it is not even clear that the government will recognize an inefficient standard in practice.278 Third, as has been amply noted in the literature on
public choice, government agencies in a position to influence the outcomes of market competition
274.
See, e.g., Joseph Farrell & Carl Shapiro, Standard Setting in High Definition Television, 8
Brookings Papers on Econ. Activity 1 (1992); Nicholas Negroponte, Being Digital 37-40 (1995).
275
See Joel Brinkley, U.S. and Europe in Battle Over Digital TV, N.Y. Times, Aug. 25, 1997, at C2.
276.
This arguably has happened in the HDTV context, where the standard (set five years ago) has not
proven to be the best technology in the new computer-driven market. See Caruso, supra note 271, at C5.
277.
Libicki studied government efforts to support seven sets of standards, and concluded that the
government’s efforts were generally unsuccessful, in part because “government is ponderous; it gets under way
slowly and once a course is set plods on, well after everyone else may have taken a different path.” Martin C.
Libicki, Standards: The Rough Road to the Common Byte, in Standards for Information Infrastructure, supra note
154, at 35, 75.
278.
Ilene Knable Gotts and Alan Rutenberg make a related point—that a government-mandated
compatibility standard may reduce the incentives for innovation in the design of interfaces. Ilene Knable Gotts &
Alan D. Rutenberg, Navigating the Global Information Superhighway: A Bumpy Road Lies Ahead, 8 Harv. J.L.
& Tech. 275, 320 (1995).
[Vol.
YEAR]
SHORT TITLE
99
are highly susceptible to "capture" by private entities with an interest in the outcome.279 Thus,
there is no guarantee that a government standard-setting body will act in the public interest, even if
it is possible for them to discern what in fact that interest is.280 Examples of both of these problems
can be found as far back as the eighteenth century, when the British government passed the Longitude Act of 1714. This Act offered a substantial reward to anyone who could determine precisely
how to measure longitude. Two solutions were proposed. Unfortunately, the best technical solution was not the politically popular one. Even more unfortunately, the government standards board
took nearly forty years to decide between the competing approaches.281
Finally, someone must decide which government sets the standard. Since we lack an effective
meta-government, every government will have to consider for itself whether it should be in the
279.
See generally Gabriel Kolko, Railroads and Regulation (1965); Theodore Lowi, The End of Liberalism
(1969); R. Fellmeth, The Interstate Commission Omission: The Public Interest and the ICC (1970).
Alternatively, even a government that is not acting at the behest of any private group may also hold its standard
hostage to extraneous policy concerns.
The government’s repeated efforts to force the use of key-escrow
cryptography, for example, have resulted in its refusal to allow the industry to converge on an efficient international
standard for encryption. On this issue, see, for example, A. Michael Froomkin, The Metaphor is the Key:
Cryptography, the Clipper Chip, and the Constitution, 143 U. Pa. L. Rev. 709, 788-89 (1995) and Brock N.
Meeks, Still Sucks: Clipper III, Wired, Aug. 1996, at 37.
280.
Cf. McGowan & Lemley, supra note 163, at 315-22 (making an analogous argument with respect to
government restrictions on trade). Once again, the argument can be made that the HDTV standard illustrates this
problem. Even the current head of the F.C.C., which set the standard, appears to view it as “a creation of the
broadcasting industry” designed to promote their interests. Caruso, supra note 271, at C5.
281
For an entertaining discussion of these events, see Dava Sobel, Longitude: The True Story of a Lone
Genius Who Solved the Greatest Scientific Problem of His Time (1995).
++100
CALIFORNIA LAW REVIEW
business of setting a particular standard. This raises a number of potential problems. A large national government might wrongly impose uniform regulation on a variety of sectors, only part of
which exhibit network effects. Alternatively, a smaller government might regulate only part of a
network that is broader than its territorial reach.282 Worse, different governments (or even different
agencies within the same government) might set inconsistent standards, driving a networks market
away from uniformity rather than toward it. The traffic example starkly presents the question of
choice of government. Because there are learning costs associated with switching driving lanes,
the optimal solution would be to have a uniform world rule regarding lane choice. Unfortunately,
different governments have locked in different rules (left side in England, right side in the United
States), and changing either rule would impose a significant cost.283
We should also be concerned about government-set standards where the need for uniformity is
not completely clear. As we have seen, it is easy to confuse network-driven markets for markets
that merely exhibit certain economies of scale. In many industries, network effects may be sufficient only to drive a part of the market toward standardization. In those circumstances, it would be
unwise to compel uniformity.284 This concern is really a combination of two potential problems:
that the government will think a standard is necessary when it isn’t; and that it will expand the
282
This appears to be behind the mass of overlapping and often inconsistent state regulations of the
Internet that have begun to appear.
283.
For example, the Russian government in the Czarist days deliberately built its railroad gauges at a
different width than the prevailing European gauge, allegedly to make invasion from Europe more difficult. The
difference in gauges persisted for a considerable period of time. See generally Less Car Load Lots v. Pennsylvania
R. Co., 10 F. Supp. 642, 647 (S.D.N.Y. 1935) (alluding to the different gauges then in force).
284.
See Richard Nelson, in Farrell & Shapiro, supra note 274, at 78-79 (expressing concern that a
government-set HDTV standard would end up displacing existing television standards unnecessarily).
[Vol.
YEAR]
SHORT TITLE
101
standard too far in product space (rather than geographic space), forcing consumers who would
otherwise have been better off buying a separate product to buy the standard product or nothing.
We should distinguish government standard-setting from another, more innocuous, form of
government activity. The government may also act from time to time as a market participant in a
way that affects standards. This may be either inadvertent or a deliberate effort to use purchasing
power to back a single standard.285 The government will affect standard choice by its purchasing
decisions, whether it wants to or not. Any large purchaser of, say, Microsoft’s operating system
will help reinforce the network effects in that system. But even where the government acts deliberately to support (or undermine) a standard, such action does not raise the same sorts of concerns as
mandatory standards, at least where the government lacks monopsony power. The market can and
will ignore government efforts at leadership if its interests lie elsewhere. Consider the case of the
metric system in the United States. The government sought for a short period in the 1970s to promote national acceptance of the metric system of measurement, the system uniformly used in the
rest of the world. Government efforts at education and at “priming” the transition by switching
government signs and publications to metric failed to tip the market, however. The government
soon abandoned these efforts, and ever since has remained a force in favor of the English system
by publishing information (such as speed limit and mileage signs) using only that system. Ironically, at about the time the government stopped supporting the metric system for general use, the
tool and automotive industries switched en masse to the metric system, apparently driven by the
growing internationalization of competition in their industries. The government is unquestionably
285.
See Howard S. Dakoff, Note, The Clipper Chip Proposal: Deciphering the Unfounded Fears that are
Wrongfully Derailing its Implementation, 29 J. Marshall L. Rev. 475, 482-84 (1996) (identifying one case where
the government has done so).
++102
CALIFORNIA LAW REVIEW
a player in setting this standard, but it is equally clear that its voice is not determinative, and that in
fact the naturally developing standard will vary from time to time and from industry to industry,
depending on the dynamics of each.
D. Telephony
“How can an Act that says ‘shall’ 2,036 times be deregulatory?”
Joseph Farrell286
Network effects arguments have had no trouble taking root in the telephone industry.287 Indeed, telephony is itself the paradigmatic example of a “pure” network, about which even skeptics
like Liebowitz and Margolis concede that the network effects story has some validity.288 The very
point of the telephone network is to connect one person to other people. It is no surprise, therefore, that the efficient number of telephone networks worldwide is one. We are all better off connected to the same phone network than we would be connected to different phone networks.289
One can conceive of this either as a cost advantage (denser phone networks have lower per-unit
costs than sparse ones, and the greater your share of a market, the more dense your network will
286.
Joseph Farrell, Creating Local Competition, 49 Fed. Comm. L.J. 201, 211 (1997) (referring to the
Telecommunications Deregulation Act of 1996).
287.
See generally Glen O. Robinson, The “New” Communications Act: A Second Opinion, 29 Conn.
L. Rev. 289, 323-24 (1996) (articulating the standard network effects explanation).
288.
Liebowitz & Margolis, Uncommon Tragedy, supra note 5, at 139-40.
289.
See, e.g., Farrell, supra note 286, at 203.
[Vol.
YEAR]
SHORT TITLE
103
be relative to competitors) or as a bandwagon network effect (everyone wants to be connected to
the network to which everyone else is connected). In fact it is both.290
Regulators were quick to recognize the cost advantage of market share in telephone networks.
This advantage corresponded to a well-known, if rare, economic beast known as the natural monopoly. A natural monopoly is simply a market in which average total cost declines over the entire
range of market demand, so that it is most efficient for one producer to serve the entire market.291
[insert Figure 1 here]
290.
One should, however, distinguish between arguments for so-called “universal service” by subsidizing
local telephone rates and arguments for including everyone who would buy a phone in an unregulated market within
the same network. As Farrell notes, universal service is not a necessary consequence of network effects. Id. at 21213. See also Robinson, supra note 287, at 324-25 (making the important point that network benefits decline at the
margins, so that it is an empirical question whether adding every last person to the network via subsidy is of net
social benefit). We do not, after all, have a “universal service” requirement for software operating systems, under
which everyone in the country is entitled to buy Microsoft Windows at less than cost. Accord id. at 323. And while
a subsidy from outside the telephone industry might have the salutary effect of artificially expanding the number of
telephone users, to the benefit of all network members, it is hard to see what can be accomplished by an internal
subsidy in which some network members support others. Such an internal subsidy might serve as a form of price
discrimination, placing an increased burden on those with a greater presumed ability to pay. Whether this would be a
good or bad thing depends on one’s perspective, however, and on the incentives it will create for arbitrage. See, e.g.,
Richard Markovits, A Constructive Critique of the Traditional Definition and Use of the Concept of “The Effect of a
Choice on Allocative (Economic) Efficiency”: Why the Kaldor-Hicks Test, the Coase Theorem, and Virtually All
Law-and-Economics Welfare Arguments are Wrong, 1993 U. Ill. L. Rev. 485 (exploring arbitrage in response to
price discrimination).
291.
See, e.g., Pierce, supra note 13, at 12-15.
++104
CALIFORNIA LAW REVIEW
To the extent that anyone thought about network effects back then, such effects seemed to bolster
the natural monopoly story, since network effects predicted that a single telephone network would
emerge from the competition between different networks at the turn of the century. Indeed one
network did emerge victorious from this standards competition—Bell’s. At the time, the accepted
thing to do when one encountered a natural monopoly was to regulate it, because competition was
presumed unable to provide long-run benefits in such a market, and because without some form of
regulation the monopolist would charge a supracompetitive price.292
If done right, regulation
would in theory set the minimum efficient price necessary for the monopolist to recover its operating costs and a reasonable return on its capital investment.293 Of course, one of the major lessons
of economics in the last fifty years is that regulation is almost never done right, for a variety of reasons: imperfect information, industry “capture” of regulators, and so on.294
In hindsight, it should perhaps have been obvious that network effects did not dictate a telephone network run by a single firm as a regulated natural monopoly. Rather, as in the intellectual
property and antitrust cases we have considered so far, network effects merely dictated that one
292.
See id. at 14; Paul Samuelson & William Nordhaus, Economics 911 (12th ed. 1985) (discussing the
economics of natural monopoly).
293.
The historic revenue requirement for regulated monopoly firms is R = O + (K-d)(r), where the revenue
requirement R is set at the operating expenses including depreciation of capital equipment (O) plus a rate of return (r)
on the current rate base (capital investment (K) less depreciation (d)). E.g., Pierce, supra note 13, at 51; Sidney
Shapiro & Joseph Tomain, Regulatory Law and Policy 198 (1993). Of course, there is significant dispute on how
that price should be set in the telecommunications industry. For one view, see J. Gregory Sidak & Daniel F.
Spulber, The Tragedy of the Telecommons: Government Pricing of Unbundled Network Elements Under the
Telecommunications Act of 1996, 97 Colum. L. Rev. 1081 (1997).
[Vol.
YEAR]
SHORT TITLE
105
network was the efficient outcome. How many different firms participated in that one network
was purely a question of interoperability—competition could perhaps have been regulated rather
than excluded entirely.295 Nonetheless, for most of this century, the Federal Communications
Commission took its mandate to be the exclusion of competition from the telephone market, and
the regulation of AT&T as a monopoly provider of telephone services and equipment.296 Beginning
in the late 1960s with the Carterfone decision,297 the FCC grudgingly began to allow competition
into first the equipment and then the long-distance segments of the market. Competition in those
areas—and the principle of nondiscriminatory interconnection—were cemented in the consent decree breaking up AT&T.298 Once the single phone network was divided into seven “regional Bell
operating companies” (plus some independents) in charge of local phone service, and a potentially
unlimited number of long-distance carriers, it was evident to all that interconnection was at the
heart of the phone system.
294.
See generally Pierce, supra note 13, at 173-211.
295.
Indeed, some have argued that competition was a greater spur to network growth than a standard owned
by a single company. See Robinson, supra note 287, at 322; Richard Gabel, The Early Competitive Era in
Telephone Communication: 1893-1920, 34 Law & Contemp. Probs. 340, 344-45 (1969). On the other hand, if
interconnection was technically more difficult in the past, that might have made the regulatory approach problematic.
296.
In re Policy and Rule Concerning Rates for Dominant Carriers, 4 FCC Rcd. 2873, 2882-88 (1989)
(reviewing this history); see also Farrell, supra note 286, at 204-06.
297.
In re Use of the Carterfone Device in Message Toll Telephone Service, 13 F.C.C.2d 420 (1968). See
also Hush-A-Phone Corp. v. FCC, 238 F.2d 266, 269 (D.C. Cir. 1956) (allowing attachment of non-AT&T
equipment which did not affect the phone or the network).
298.
United States v. Western Electric, 552 F. Supp. 131, 227 (D.D.C. 1982), aff’d sub nom. Maryland v.
United States, 460 U.S. 1001 (1983).
++106
CALIFORNIA LAW REVIEW
The problem then became how to ensure effective interconnection. In part, this is a technical
and accounting problem—what standards should be set to ensure that different companies’ phone
systems work together, and who should foot the bill.299 But it is also an antitrust problem involving incentives to discriminate. The Antitrust Division took the position in AT&T that it could not
prevent discrimination in interconnection unless the local telephone monopolists were “walled off”
from the competitive long-distance sector.300 This solution is rather drastic from an antitrust perspective—no one is seriously suggesting that we break up Microsoft, for example (or any other
vertically integrated company, for that matter).
More recently, Congress tried a different approach. In the 1996 Telecommunications Reform
Act,301 Congress agreed to break down the wall separating local from long-distance providers, on
the condition that the local telephone market itself be opened to competition. The theory is that local telephone providers cannot effectively discriminate in favor of a particular long-distance com-
299.
The much-celebrated convergence of various communications media, including conventional
telephony, wireless telephony and coaxial cable, complicates the technical problem, because each should ideally
interconnect to the other. So too, does the rise of Internet telephony, which presents a rather different set of
interconnection issues. See Dennis Moore, Internet Telephony, 76 Tex. L. Rev. 183 (1997); Robert Cannon, The
Internet at the FCC: Cybernauts vs. Ma Bell, http://www.cais.net/cannon/ (visited June 11, 1997); European
Telecom Standards Body to Work on Standard for Internet Telephony, 2 Electronic Info. Pol. & L. Rpt. 520, 520
(May 16, 1997).
300.
See Roger G. Noll & Bruce M. Owen, The Anticompetitive Uses of Regulation: United States v.
AT&T, in The Antitrust Revolution 290 (Kwoka & White, eds., 1989).
Farrell refers to this approach as
“quarantining the monopoly lest it infect the competitive segments.” Farrell, supra note 286, at 207.
301.
Pub. L. No. 104-104, 110 Stat. 56 (1996).
[Vol.
YEAR]
SHORT TITLE
107
pany if local telephone competition constrains them.302 This may or may not be true; it remains to
be seen whether individual consumers will switch their local service provider to avoid longdistance discrimination, at least where the discrimination does not impose major costs on them. In
any event, it seems likely that whether or not the local market is opened to competition, the FCC
will continue to impose mandatory interconnection requirements on all the relevant players.303
Network effects arguments are virtually irrefutable in telephony. As history demonstrates,
however, accepting that fact does not dictate the course the industry must take.
The current
course—opening the network to competition by compelling comparably efficient interconnection
between all players—seems a logical one, given what we know of network effects. 304 But requiring interconnection may force the government to remain a player in this market for the foreseeable
future, since at least some market players may have an incentive not to interconnect with all comers
on nondiscriminatory terms. As Farrell contends, therefore, “for true sharing of network externalities, and for true sharing of economies of density, some intervention in the bargaining process
is likely needed.”305
302.
See Farrell, supra note 286, at 207-08.
303.
See 47 U.S.C. § 255-56 (1996) (imposing interconnection obligation pursuant to regulations set by
the FCC). It has also continued to restrict the power of the RBOCs to enter other markets pending the development
of genuine competition. See Leslie Cauley, Genuine Competition In Local Phone Service Is a Long Way Off, Wall
St. J., Dec. 15, 1997, at A1; Leslie Cauley, Baby Bells Remain Bystanders in Phone Takeover Wars, Wall St. J.,
Oct. 16, 1997, at B4.
304.
See supra notes 152-162 and accompanying text (discussing interoperability as a solution to
monopoly power created by network effects).
305.
Farrell, supra note 286, at 211 (suggesting possible government regulations to ensure interconnection
at efficient prices).
++108
CALIFORNIA LAW REVIEW
E. Internet Governance
The Internet, like the telephone network, exhibits a strong form of network effects—the network is the product in a very real sense.306 Given our experience with the telephone network, and
the parallels between the two, one might expect the Internet to exhibit the same market structure.
But it does not. The Internet looks nothing like the old Bell System, with a single regulated corporation in charge of all the connections. In fact, it doesn’t even look like the new, streamlined
model of telephone competition, in which various large network owners will be forced to interconnect on government-set (and government-enforced) terms. Rather, the Internet appears (at least at
first glance) to be an example of working anarchy—it consists of millions of different entities
around the globe, public and private, which connect together in a patchwork network of uncertain
provenance.
How can this be? The answer is twofold. First, the Internet is not composed entirely of its
own set of lines connecting one Internet user to the next, though some Internet-specific “backbone”
wires do exist. Rather, the Internet piggybacks on existing communications technology, notably
the wire telephone network, using available space to send distributed packets of information from
place to place.307 As a result, no one can fairly be said to have “built” the Internet in the physical
sense, though of course a number of people built early pieces of it, and United States government
306.
See Lemley, Internet Standardization, supra note 35, at 1044-45.
307.
See generally Joshua Eddings, How the Internet Works (1994).
[Vol.
YEAR]
SHORT TITLE
109
agencies such as the National Science Foundation and the Defense Advanced Research Projects
Agency contributed parts of the network backbone. 308
Second, and more important, the Internet itself is nothing more than a relatively simple set of
computer protocols (commonly called TCP/IP today) governing the interchange of data. In other
words, what we think of as “the Internet” is really only a published, nonproprietary interface standard. 309 Anyone who uses the standard to transmit data from her computer is “on” the Internet;
anyone who doesn’t use the standard is not. The success of the Internet is due largely to its spectacular interoperability. It did not drive out its competitors in the “computer networking” market so
much as subsume them.310 In the mid-1980s, one could participate in computer networking by
joining one of approximately 50,000 bulletin board systems (or BBS’s), or one of the fledgling
on-line service providers like Prodigy or Compuserve, or (depending on where one worked) one
of the private networks of military or academic computers. Each of these computer networks was
largely incompatible with the others, with the result that joining a bulletin board allowed you to
communicate only with other members of that bulletin board. Interconnection protocols, beginning
with Usenet and SMTP, allowed messages to be transferred between different groups of networked computers. As the communications technology between networks became more seamless,
people began to think of themselves as on the Internet itself, rather than connected to a private
308.
For an entertaining discussion of the early development of the Internet, see Katie Hafner, Where
Wizards Stay Up Late (1996).
See also Barry M. Leiner et al., A Brief History of the Internet
(http://www.isoc.org/internet-history/#fricc) (visited Feb. 12, 1997).
309.
See Leiner et al., supra note 308. David G. Post, Anarchy, State and the Internet: An Essay on Law-
Making in Cyberspace, 1995 J. Online L. art 3, at ¶¶ 12-13.
310.
For a good legal description of the Internet, see ACLU v. Reno, 929 F. Supp. 824, 830-31 (E.D. Pa.
1996), aff’d, 117 S.Ct. 2329 (1997).
++110
CALIFORNIA LAW REVIEW
computer networking group which could itself exchange data with other private groups. From the
perspective of network economic effects, the Internet is a tremendous success story because it allows different and often incompatible computer systems to communicate with each other, expanding the size of the network without requiring purchase from the standards owner.
Things are not quite as simple as this, however. TCP/IP does not do everything automatically
and without supervision, any more than do the stock exchanges, often cited as the most efficient of
“unregulated” markets.311 Rather, there are a wide variety of rulemaking groups that enforce standards on the Internet. For example, technical standards (including updates or changes to the
TCP/IP protocol) are set by the Internet Engineering Task Force (IETF), a voluntary body with no
“official” authority over the Internet beyond the willingness of Internet users to adopt the changes it
proposes.312 Further “rules” are created by the code written both into the Internet protocols and into
311.
See Ronald H. Coase, The Institutional Structure of Production, in Essays on Economics and
Economists 112 (1994) (“[S]tock and produce exchanges are often used by economists as examples of perfect or nearperfect competition. But these exchanges regulate in great detail the activities of traders (and this quite apart from
whatever public regulation there may be).”).
312.
For a discussion of standard-setting in the all-volunteer IETF, see A. Michael Froomkin, A Model of
International Law and Society 16-22 (working paper 1997); A. Michael Froomkin, The Internet as a Source of
Regulatory Arbitrage, in Borders in Cyberspace [recommend pincite] (MIT Press 1997); Lewis C. Lee & J.
Scott Davidson, Intellectual Property for the Internet 180-82 (1997); William Lehr, Compatibility Standards and
Interoperability: Lessons from the Internet, in Standards for Information Infrastructure, supra note 154, at 131-37;
Paulina Borsook, How Anarchy Works, 3.10 Wired at 110 (Oct. 1995). It is interesting, and for our purposes
perhaps instructive, that the quasi-”official” body designing computer and communications standards, the
International Standards Organization (ISO), during the 1980s refused to accept TCP/IP as a networking standard.
[Vol.
YEAR]
SHORT TITLE
111
privately owned programs, like Netscape Navigator, Microsoft Internet Explorer, and Sun’s Java,
that operate on the Internet.313
For our purposes, though, the best example of the role of network effects in Internet governance has to do with domain names. InterNIC (and the private company Network Solutions, Inc.)
set rules for addressing on the Internet, mapping the “domain names” commonly used to identify
individuals and corporations on the Net to the numeric IP address that is actually registered to a
particular server.314 For addressing to work, someone must maintain a list of valid IP addresses
Instead, the ISO offered its own standard, called Open Systems Interconnection (OSI). OSI failed to replace the basic
TCP/IP standard, which continues to govern the Internet. See Froomkin, Model of Interational Law, supra, at 15.
313.
An important and growing body of legal scholarship has focused attention on the role of “code” in
setting quasi-legal rules on the Net. See, e.g., Lawrence Lessig, Reading the Constitution in Cyberspace, 45 Emory
L.J. 869 (1996) [hereinafter Lessig, Constitution in Cyberspace]; Lawrence Lessig, supra note 203; Lawrence
Lessig, The Constitution of Code: Limitations on Choice-Based Critiques of Cyberspace Regulation, 5 CommLaw
Conspectus 181 (1997); David Post, Bargaining in the Shadow of the Code: File Caching, Copyright, and Contracts
Evolving in Cyberspace (working paper 1997); Reidenberg, supra note 182. For a concrete example of “code-based
regulation,” see the Digital Telephony Act of 1994, Pub. L. No. 103-414, 108 Stat. 4279; see also Susan Freiwald,
Uncertain Privacy, 69 S. Calif. L. Rev. 949 (1996).
314.
The interaction between InterNIC’s “first-come, first-serve” rule for allocating Internet domain names
and trademark law has been explored elsewhere. See, e.g., Dan L. Burk, Trademarks Along the Infobahn: A First
Look at the Emerging Law of Cybermarks, 1 Rich. J.L. & Tech. 1 (1995); Gary W. Hamilton, Trademarks on the
Internet: Confusion, Collusion, or Dilution?, 4 Tex. Intell. Prop. L.J. 1 (1995); David J. Loundy, A Primer on
Trademark Law and Internet Addresses, 15 J. Marshall J. Comp. & Info. L. 465 (1997); James West Marcovitz,
[email protected]—“Owning a Bitchin’” Corporate Trademark as an Internet Address—Infringement?, 17
Cardozo L. Rev. 85 (1995); Ira S. Nathenson, Showdown at the Domain Name Corral: Property Rights and Person
Jurisdiction Over Squatters, Poachers and Other Parasites, 58 U. Pitt. L. Rev. 911 (1997); Carl Oppedahl, Remedies
++112
CALIFORNIA LAW REVIEW
and their aliases. Further, everyone on the Internet must work from that list, so that a user who
types in “www.ibm.com” will find the computer identified with that alias. Network Solutions
(NSI) has performed this function for several years now with respect to certain of the international
top level domains, or iTLDs, like “.com”.315 The list of names and matching IP addresses are entered into a series of “root domain name servers” or DNS’s that are run by volunteers.316 The servers are themselves updated regularly. If your name is on the list, you can be found on the Internet.
If not, anyone who types that name into their browser or mail program will not be able to reach
you. 317
Obviously, the DNS servers are central to the functioning of the Internet. And if you want
your name to have a DNS entry, the only obvious way to get it is to request that NSI or one of the
in Domain Name Lawsuits: How is a Domain Name Like a Cow?, 15 J, Marshall J. Comp. & Info. L. 437
(1997). See also http://www.ll.georgetown.edu/lc/internic/domain1.html (cataloguing information about domain
name trademark disputes). On an effort to challenge those rules, see Roderick Simpson, Dueling Domains, Wired,
Aug. 1996, at 64.
315.
There are also national TLDs for each country, ending with the two-letter code for that country.
Administration of these national TLDs is generally in the hands of entities within each country, though of course
national TLDs must also be included in the root DNS if they are to be accessible to people elsewhere on the Internet.
It is NSI which coordinates the inclusion of the relevant lists.
316
Rebecca Quick, Is the Internet Outgrowing Its Volunteer Traffic Cops? Wall St. J., Sept. 12, 1997, at
B6.
317.
The recent corruption of data in root DNS servers graphically illustrated this problem, bringing down
most Internet addressing for several hours in July of 1997. See John Markoff, Ignored Warning Leads to Chaos on
the Internet, N.Y. Times, July 18, 1997, at A1; David G. Post, Breaking Up the Domain Name Monopoly,
Recorder, Sept. 25, 1997, at 4.
[Vol.
YEAR]
SHORT TITLE
113
national registrars it works with put it there.318 But NSI is a private entity; it is not at all clear what
authority it has to “run” the Internet, or indeed who might be able to give it that authority. NSI
took over operation of the domain name registration system based on a contract from the U.S. National Science Foundation. That contract will expire in 1998, and NSF has already indicated that it
will not renew the contract.319
So what happens to the database of names at the heart of the DNS system? Consider several
possible outcomes. First, NSI may decide to keep administering the DNS system, and the courts
may decide it has a right to control the database of names. If this occurs, control over the network
will effectively have been placed in the hands of a single private entity, which will become the de
facto standard-setter. In theory, NSI could be displaced from this position by market competition—someone else could set up a competing, incompatible domain name server, and if enough
users of the Internet (and therefore of the NSI DNS system) switched to this incompatible system,
the owners of the new server would become the new market leader. In practice, network effects
make this outcome unlikely, because a “competing Internet” that few people are using will not be
318.
“If you want to attach your network to the Internet, but you don’t like NSI’s policies, for whatever
reason, you quickly learn that NSI is the only game in town.” Copyright Protection on the Internet, Hearings
Before the Subcommittee on Courts and Intellectual Property of the House Committee on the Judiciary on H.R.
2441, 104th Cong., 2d Sess. (1996) (statement of Catherine Simmons-Gill on behalf of the International Trademark
Association). In Larry Lessig’s parlance, this is more of a “code” restriction than a policy matter. If the root DNS
name servers are not configured to “recognize” a TLD (like “.web”), they simply won’t match it with an IP address,
and messages to that domain will not be delivered. Period.
319.
See David S. Hilzenrath, Network Solutions Dropped as Registrar of Internet Domains, Wash. Post,
April 27, 1997, at E1; NSF Won’t Renew Network Solutions’ Contract, Searcher, June 1997, at 40.
++114
CALIFORNIA LAW REVIEW
attractive to those already on the current Internet. 320 Indeed, some halting efforts toward such an
alternate DNS system have so far been unsuccessful.321 While NSI’s control over DNS might also
be displaced by hacking,322 any such unauthorized access to the DNS servers is likely to be illegal.323
320.
Unlikely is not impossible, however. There are some reasons to believe that even here, network
effects might not prevent effective competition between standards. First, the switch to a new domain name system
need not be a complex one. NSI cannot claim to own the basic protocols that govern the Internet. It might be
relatively straightforward, therefore, for a concerted group of large Internet users to switch their allegiance in a public
way, causing others to follow suit. Second, and more important, it might be possible to run a new DNS system
alongside the existing one, so that a company could be on both systems at once. If this is feasible, IBM could be
accessed through ibm.com via NSI, and through a different (or conceivably even the same) domain name on a
different system. Which system a user used would depend on how she accessed the Net. Network effects are
significantly alleviated to the extent that users can simultaneously use more than one standard, as we have seen. See
supra notes 137-151 and accompanying text.
321.
For example, both Alternic and a DNS alternative called eDNS exist, but neither routes a significant
amount of traffic because most people do not have their routers configured to take instruction from these
“alternative” DNS servers. For more on these alternatives, see Neal J. Friedman & Kevin Siebert, The Name is Not
Always the Same, 20 Seattle U. L. Rev. 631, 657-61 (1997).
322.
Such a hack would involve spoofing NSI’s identity to convince the DNS server to accept updated
information provided by an unauthorized third party. For a technical discussion of the possibility of DNS entry
hacking, see Secure Networks Inc., BIND Vulnerabilities and Solutions, Security Advisory April 22, 1997
(ftp://ftp.secnet.com/advisories/SNI-12.BIND.advisory) (visited July 1, 1997). Such a BIND hack was perpetrated in
July by Alternic, a company challenging NSI’s exclusive authority over TLDs. Alternic may now face criminal
prosecution for its actions. See Todd Wallack, Net Domain Name Squabble Takes Unexpected Route, Network
World, July 28, 1997, at 10.
[Vol.
YEAR]
SHORT TITLE
115
A second possibility is that NSI will get to keep its control over the DNS database, but that
some legal constraint will be imposed on its discretion in running the DNS system. Some such
constraints are obvious—for example, the strong weight of authority holds that NSI’s domain
name registration policy must yield to the contrary dictates of trademark law.324 But there have also
been hints of a more fundamental governmental role in compelling access to NSI’s domain name
servers. Indeed, litigation on this issue has already begun. In a complaint filed in March 1997
against NSI, a company called PGP Media alleged that NSI’s failure to incorporate PGP-registered
323.
See, e.g., 18 U.S.C. § 1030 (prohibiting unauthorized access to network computers).
324.
See. e,g., Comp Examiner Agency v. Juris, Inc., 1996 U.S. Dist. LEXIS 20259 (C.D. Cal. April 25,
1996) (injunction vs. direct competitor); Actmedia, Inc. v. Active Media Int’l, 1996 US Dist. LEXIS 20814 (N.D.
Ill. July 17, 1996) (same); Cardservice Int’l v. McGee, 950 F. Supp. 737 (E.D. Va. 1997) (same); Planned
Parenthood Federation v. Bucci, 42 USPQ2d 1430 (S.D.N.Y. 1997) (anti-abortion activist can’t use Planned
Parenthood name); Intermatic v. Toeppen, 947 F. Supp. 1227 (ND Ill 1996) (same dilution analysis); Panavision v.
Toeppen, 945 F. Supp. 1296 (C.D. Cal. 1996) (Toeppen’s interactive map enjoined as dilution, but not as trademark
infringement; reselling domain name is “commercial use”); Hasbro, Inc. v. Internet Entertainment Group, 40
USPQ2d 1479 (W.D. Wash 1996) (adult site dilutes famous name for children’s game); Toys’r’Us v. Akkaoui, 40
USPQ2d 1836 (N.D. Cal. 1996) (dilution of “‘r Us” mark by defendant’s “adultsrus” domain name); Inset Systems
v. Instruction Set, Inc., 937 F. Supp. 161 (D. Conn. 1996) (dictum stating that use of a trademark as a domain
name may cause confusion in the marketplace). But see Giacalone v. Network Solutions, 1996 U.S. Dist. LEXIS
20807 (N.D. Cal. June 14, 1996) (injunction against NSI placing a registered domain name on hold at the request of
a trademark owner). Cf. DEC v. Altavista Technology, 960 F. Supp. 456 (D. Mass. 1997) (injunction vs. ATI’s
use of “altavista” for services, even though it was licensed by DEC to use altavista.com). These cases in the
aggregate clearly establish that it is trademark law, not NSI’s “first-come, first-serve” registration policy, that will
determine who owns a domain name where the two policies are in conflict.
++116
CALIFORNIA LAW REVIEW
domains violated the antitrust laws.325 The claim is that NSI’s configuration file “is the central (and
essential) technical bottle-neck facility for the Domain Name Registration Market,”326 and the plaintiffs seek “extremely limited and narrowly tailored injunctive relief to compel NSI to add reference
in the Configuration File on the NSI Root Nameservers . . . so that PGP may compete with NSI in
the Domain Name Registration Market for Domain Name registrations.”327 There are a number of
potential problems with such a claim, including a possible antitrust immunity defense based on
state action328 and the general reluctance of courts to declare privately held facilities to be essential.329 But that the claim was filed at all pays tribute to the growing confusion over precisely who
should have authority to “issue” domain names.330 Competitors could also seek to compel access to
NSI’s database of DNS entries on an intellectual property theory.331
325.
PGP Media, Inc. v. Network Solutions, Inc., No. 97 Civ. 1946 (RPP) (S.D.N.Y. filed March 20,
1997), available at http://www.jmls.edu/cyber/cases/pgpm-c.html (visited March 24, 1997). See generally Alexander
Gigante, Blackhole in Cyberspace: The Legal Void in the Internet, 15 J. Marshall J. Comp. & Info. L. 413, 430
(1997) (suggesting that NSI as a private actor might be subject to antitrust liability).
326.
PGP Media, No. 97 Civ. 1946 at 1.
327.
Id.
328.
For a detailed discussion of the state action doctrine, see McGowan & Lemley, supra note 163, at 315-
60. PGP Media did anticipate this defense, specifically alleging that NSI is no longer effectively supervised by
NSF. PGP Media, No. 97 Civ. 1946, at __ [recommended pincite]. Once the NSF-NSI contract expires, of
course, this claim will be even stronger.
329.
See supra notes 121-122 and sources cited therein (discussing the essential facilities doctrine).
330.
Private suits are not the only means of raising antitrust claims. In July of 1997, NSI disclosed in an
SEC filing that the Antitrust Division of the Department of Justice was investigating whether NSI had violated the
[Vol.
YEAR]
SHORT TITLE
117
Finally, the government could decide to remove NSI’s control over domain name registration
entirely, replacing it with either a de novo open system of private registration, or even a system of
government registration. In February 1997, the International Ad Hoc Committee (IAHC), a nongovernmental Internet advisory group set up to study the issue, recommended the creation of seven
new “generic top-level domains (gTLDs),” to be administered on a global basis by “multiple competing registrars” overseen by a (private) Council of Registrars and a (private) DNS Policy Oversight Committee.332 The international nature of this approach is an important indication that one
cannot merely say “the government” will establish a new policy. While the IAHC report was prepared under the informal auspices of the World Intellectual Property Organization, and the
“Memorandum of Understanding” it established333 is deposited with the International Telecommunications Union, no governmental or international legal authority stands behind the proposal, a fact
which has upset both the United States and the European Union.334 Indeed, the Memorandum of
antitrust laws in its administration of the domain name registration system. See Jennifer B. Lucas, Domain Name
Practices Are Subject of DOJ Antitrust Investigation, Elec. Info. Pol. & L. Rpt., July 18, 1997, at 748-49.
331
NSI has apparently agreed to allow access to its database, but not to its software for operating the
database. See Kelly Flaherty, NSI Focus on Internet Talks in Congress, Recorder, Oct. 2, 1997, at 1, 4.
332.
Final Report of the International Ad Hoc Committee: Recommendations for Administration and
Management of gTLDs, available at http://www.iahc.org/draft-iahc-recommend-00.html (visited June 19, 1997).
The IAHC report declares that iTLDs are within its “purview,” but cites no authority.
333.
Establishment of a Memorandum of Understanding on the Generic Top Level Domain Name Space of
the Internet Domain Name System (gTLD-MoU), available at http://www.iahc.org/gTLD-MoU.html (visited June
19, 1997).
334.
See David Loundy, E-Law: International Intrigue Meets Internet Domain Name System, Cyberspace
Lawyer, June 1997, at 14-16.
++118
CALIFORNIA LAW REVIEW
Understanding declares that it is agreed to by “The Internet Community,”335 an ethereal entity if
ever there was one. NSI probably has no enforceable legal obligation to comply with a dictate
from such a body, though of course it might well end up being incorporated into national or international law.336 Alternatively, whatever private group claims authority over Internet domain names
is undoubtedly subject to existing national laws, at least if the country applying those laws has jurisdiction.337
From an economic perspective, there is clearly more social value in a shared than in a monopolistic top-level domain system. Internet domain names are valuable commodities; one generic
domain name was recently sold for over $100,000.338 NSI currently charges only $100 per year
per domain name, up from $0 a few years ago. But an NSI freed of governmental constraints on
335.
Id.
336.
For a discussion of this problem, and proposals for national or international legislation along these
lines, see Gigante, supra note 325, at 426-29. See also David W. Maher, Trademark Law on the Internet—Will it
Scale? The Challenge to Develop International Trademark Law, 16 J. Marshall J. Comp. & Info. L. 1, 5 (1997)
(“The answer to the question of ‘who’s in charge?’ is ‘no one.’”).
337.
For a discussion of jurisdiction on the Internet, see Dan L. Burk, Jurisdiction in a World Without
Borders, 1 Va. J. L. & Tech. (1996) (http://www.student.virginia.edu/~vjolt/vol1/BURK.htm) (visited August 4,
1997). It is worth noting that the current administrators of global TLDs have followed a de facto policy of granting
control over “country” domains—the two-letter codes in widespread use outside the United States—to the
government of the relevant country, and not necessarily to the original delegate. See Internet Governance Not
Scaling Well, The Cook Report, Part II, http://www.cookreport.com/06.06.shtml (visited Aug. 1, 1997).
338.
See Laurie J. Flynn, Prototype Internet Name Is Sold for $100,000-Plus, N.Y. Times, May 12, 1997,
at C4 (“internet.com” domain name worth over $100,000); Rory J. Thompson, Business Deal, InformationWeek,
June 9, 1997, at 12 (“business.com” sold for record sum).
[Vol.
YEAR]
SHORT TITLE
119
its pricing policy probably has significant power to raise the price of domain names, either across
the board or by price discrimination. Competition among domain name registrars would constrain
such power, particularly if (as the IAHC report recommends) competing registrars could register
names within a single TLD. If IBM can choose to obtain rights to “ibm.com” from either NSI or
from a competitor such as PGP, it will be able to shop effectively for the lowest registration price,
and price should approach marginal cost.339
On the other hand, administering such a competitive system will be more difficult than running
the current system. Someone must control access to the DNS root servers; either one of the registrars must be trusted to give nondiscriminatory access to the servers, or some governmental body
will have to police interconnections, as is done in the telephone industry. Some provision will also
have to be made for preventing conflicting registrations, and the interaction between trademark law
and domain name registrations will become more complex. These functions don’t necessarily have
to be carried out by a government—government regulation of Internet traffic in any form makes a
lot of people nervous340 and would also subject the administrator to constitutional constraints in the
339.
By contrast, a competing top-level domain which cannot register “.com” domains might find itself
handicapped by the convenience effect of the widespread use of the .com TLD. IBM may be less likely to switch to
a competing domain name registrant if the TLD they use is an unfamiliar one, since part of the value of having a
name like “ibm.com” is that those who want to visit the site for the first time can easily guess the name.
340.
See Mike Godwin, After the CDA: Our KO’d Radio Future? The Wired Last Time (working paper
1997). See also Steve Lohr, The Internet as Commerce: Who Pays, Under What Rules?, N.Y. Times, May 12,
1997, at C1 (calling it “inevitable” that “the Federal Government will become increasingly involved in the affairs of
the Internet, even if that role is more as a referee than as a regulator”).
++120
CALIFORNIA LAW REVIEW
United States341—but they will have to be accomplished somehow if domain name registration is to
be competitive rather than proprietary. The existing system—one based on consensus—appears
unlikely to survive very long, because the cooperating parties seem fundamentally at odds..342 If
cooperation continues to characterize the DNS system, it may be for the ironic reason that the law
compels it.
Domain names are not the only fundamental part of Internet structure that turn out on inspection to be rather fragile. The system of exchanging and passing through data belonging to third
parties has to date been cooperative and, more importantly, free—a process known as “peering.”
There are some signs that that is changing. In May 1997, UUNet Technologies, a major provider
of Internet backbone connections, announced that it would charge fees to smaller Internet service
providers for passing through their messages.343 If the four other major backbone providers follow
suit, the cost of Internet access will undoubtedly rise. This is not necessarily a bad thing—the
“hidden subsidy” structure of existing Internet transmissions may well prevent efficient access
pricing and therefore discourage needed improvements in the backbones.344 But if access to the
backbone is a commodity controlled by private companies who can exclude others at will, interop-
341.
For example, a government-run domain name system would have to comply with the First
Amendment, and would have to afford registrants due process of law before revoking their rights. See Gigante, supra
note 325, at 429-30. On the other hand, a government actor would probably be free from antitrust liability under the
state action immunity doctrine. See generally McGowan & Lemley, supra note 163.
342.
See Cook Report, supra note 337, Introduction (quoting NSI’s July 3 SEC filing as reporting that
“continuing to achieve consensus may become difficult or impossible”).
343.
See Lohr, supra note 340, at C1.
344.
Hal
R.
Varian,
Differential
alfred.sims.berkeley.edu/Different/different.html>.
Pricing
and
Efficiency
(June
1996)
<http://
[Vol.
YEAR]
SHORT TITLE
121
erability concerns similar to those in telephony may lead to demands for government regulation of
the circumstances in which access can be denied.345
Finally, a word about the precise nature of the network effects at work here is in order. The
network effect grows from the positive value placed on the ability to contact other people via the
Internet, and from the access to information from a wide variety of different sources. However, it
is not the case that people on the Internet want to communicate with everyone else on the network,
just as they don’t want to receive telephone calls from everyone on the telephone network. The
death of Usenet as an effective means of communication resulted from an overabundance of participation, coupled with the lack of limits on the relevance of that participation. In this sense, then,
people want to interact with a subset of all those on the Internet, though they still may get a benefit
from expanding the pool of people with whom they can potentially interact. Similarly, people may
well want to access only a subset of information available on the Web, if that subset is filtered or
tailored in such a way as to make it more useful to them.346 “Zoning” the Net—either by setting up
private areas within it or by sectioning off entire areas by government mandate347—is not inconsistent with the idea that the Internet exhibits positive network effects. Rather, one must distinguish
between structural barriers to access set up by inconsistent market standards and contextual barriers
set up deliberately by the participants themselves.
This is not to suggest that network effects don’t have an impact on Internet governance.
Scholars sometimes argue that Internet self-governance systems are the product of choice, because
345.
Indeed, one company owner who was cut off by UUNet declared that “I am a proponent of some
regulation in the Internet as soon as possible,” calling regulation “critical to the Internet’s open and unrestricted
development.” Lohr, supra note 340, at C6.
346.
Anyone who has retrieved over 20,000 entries in an Altavista search will understand this problem.
++122
CALIFORNIA LAW REVIEW
participants could “exit”: quit the Internet in favor of an alternative computer network.348 But the
right of “exit” is illusory in a network market. The fact that everyone wants to be connected to “the
Internet”—and not some newly-created substitute—means that individual “choice” of governance
on the Internet is in fact rather weak, reduced to the level of the “choice” to take or leave a standard
form contract term.349 It also means that the Internet is far easier for governments to regulate than it
might at first appear, because access to the entire corpus of “the Internet” is so important to the
whole endeavor.
It may be that we could set up sub-groups within the Internet, and that those subgroups could
choose to govern themselves: excluding others, limiting access, and agreeing upon their own rules
of behavior. This seems to be the model Johnson & Post endorse.350 The model has its own set of
problems: to the extent there is overlap between these online enclaves, or between the enclaves and
347.
On both types of “zoning,” see Lessig, Constitution in Cyberspace, supra note 313, at 888-89.
348
E.g. David G. Post, Governing Cyberspace, 43 Wayne L. Rev. 155, 167 (1996) (“Mobility—our
ability to move unhindered into and out of these individual networks with their distinct rule-sets—is a powerful
guarantee that the resulting distribution of rules is a just one.”); Post, supra note 309, art 3, ¶ 1.
349
Indeed, the analogy is more precise than one might at first suspect, since standard form contracts
increasingly make up whatever “private governance” structure might be said to exist. See Robert L. Dunne,
Deterring Unauthorized Access to Computers: Controlling Behavior in Cyberspace Through a Contract Law
Paradigm, 35 Jurimetrics J. 1 (1994); Mark A. Lemley, Shrinkwraps in Cyberspace, 35 Jurimetrics J. 311 (1995).
350
See David R. Johnson & David G. Post, Law and Borders: The Rise of Law in Cyberspace, 48 Stan.
L. Rev. 1367 (1996); David R. Johnson & David G. Post, And How Shall the Net Be Governed?: A Meditation on
the Relative Virtues of Decentralized, Emergent Law, in Coordinating the Internet 62 (Brian Kahin & James Keller,
eds., MIT Press 1997).
[Vol.
YEAR]
SHORT TITLE
123
the real world, autonomous decisionmaking within the enclave will have effects outside it.351 Network effects add another layer of external constraint. My enclave can’t decide to forego the DNS
system unless it also decides to forego being on the Internet at all. And having opted to stay in the
network, my enclave will probably also be constrained by the legal rules that govern that network.
Certainly, I couldn’t choose “ibm.com” as my new domain name. The real problem is that the
“decentralized” law that cyberlibertarians favor for the Internet must coexist with a single set of
rules that determine the structure of the Internet itself.
F. Corporate Law
Network effects arguments recently have been used by scholars to question basic assumptions
of the contractarian paradigm that underlies much modern corporate legal theory.352 Most prominently, Michael Klausner (on his own and with Marcel Kahan) has argued that certain types of
corporate contracts—and more importantly corporate law itself—are subject to influences analogous to network effects.353 This argument poses a provocative and far-reaching instance of the
351
For example, I may want to join an enclave which has declared Microsoft’s computer programs to be
freely copyable, but Microsoft has not joined that enclave, and has legitimate grounds to object to our imposition of
this rule even “upon ourselves.”
352
On the contractarian paradigm, see generally Frank H. Easterbrook & David R. Fischel, The
Economic Structure of Corporate Law (1991).
353.
See Klausner, supra note 5. Professor Klausner continues this analysis and combines it with his
earlier work with Professor Kahan regarding event risk covenants in bond indentures in Kahan & Klausner,
Standardization, supra note 61; see also Marcel Kahan & Michael Klausner, Antitakeover Provisions in Bonds:
Bondholder Protection or Management Entrenchment, 40 UCLA L. Rev. 931 (1993) [hereinafter Kahan & Klauser,
Antitakeover]; Marcel Kahan & Michael Klausner, Path Dependence in Corporate Contracting: Increasing Returns,
++124
CALIFORNIA LAW REVIEW
challenge of adapting network theory to legal ends. Indeed, insofar as the argument contemplates
law as imbued with network effects,354 it stretches network theory further than any of the examples
we have examined thus far and makes the task of judicial adaptation even more delicate than usual.
Because Klausner's discussion of network theory in corporate law is the most extensive
treatment of the problem of adaptation in a setting of relatively weak effects,355 and because it conceives of the legal process itself as a component of network theory, we consider it at some length
here. We do not wish to overstate the case Klausner makes, however, and pause here to note that
his analysis is explicitly exploratory in nature and does not call for sweeping revisions in doctrine
or for aggressive governmental intervention, as have some proponents of network theory in other
fields.356
At a theoretical level, Klausner contends that the contractarian paradigm underlying much
modern corporate theory "fails to take account of network externalities, a type of market failure that
Herd Behavior, and Cognitive Biases, 74 Wash. U. L.Q. 347 (1996) [hereinafter Kahan & Klauser, Path
Dependence]. In addition to these works, Professor Gordon has explored the potential “positive externalities”
inhering in standardization as a basis for federalizing certain corporate governance terms. See Gordon, supra note 46,
at 1567-69; see also Lucian Arye Bebchuck, 105 Harv. L. Rev. 1435, 1493-94 (1992) (questioning strength of such
externalities); David Charny, Competition Among Jurisdictions in Formulating Corporate Law Rules: An American
Perspective On The “Race to the Bottom” in the European Communities, 32 Harv. Int’l L.J. 423, 442-45 (1991)
(arguing that positive externalities are an important benefit of standardization among jurisdictions).
354.
We use “law” broadly here, to encompass the reported decisions and statutes comprising the nominal
body of “corporate law” as well as social practices which they foster or discourage.
355.
By which we mean that each of Klausner’s effects are indirect and relatively far removed from
prototypical network goods such as telephones or interoperable compatible goods such as software or VCRs.
[Vol.
YEAR]
SHORT TITLE
125
may be uncommon in product markets, but is possibly pervasive in the market for corporate contract terms."357 This failure in turn supports his conclusion that, to the extent network effects are
significant, "the contractarian paradigm provides an incomplete account of the role of corporate law
and the process by which states compete to sell corporate charters."358 He also argues that network
theory may warrant some relatively modest change in existing corporate statutes.359
Though
Klausner rightly views network effects as "simply another type of externality that should be taken
into account in analyzing corporate contracts," he believes that "the existence of this particular type
of externality requires a basic shift in how we conceptualize corporate law—from an aid to corporate contract drafting to a standard-setting system analogous to technical standards in other
fields." 360
Klausner's work is theoretical, rigorous and ambitious; while we disagree with some of the
inferences he draws, our main purpose here is to question the degree to which network effects
cause the potentially suboptimal circumstances Klausner identifies. Consistent with our general
focus on the prospects for useful adaptation of network theory in law, our interest is in whether, as
a practical matter, network effects are likely to require modifications in either contractarian theory
or corporate law.
356
See supra notes 69-267 and accompanying text.
357.
Id. at 771.
358.
Id. at 759.
359.
Id. at 832, 837-40.
360.
Id. at 791 n.110.
++126
CALIFORNIA LAW REVIEW
1. The Contractarian Paradigm
Harkening back to fundamental Coasean insights, 361 much of modern corporate theory views
firms as webs of contracts involving shareholders, creditors, managers, and employees.362 The
contractarian view takes maximization of firm value as the goal of corporate law. Firms must
compete in capital, labor, and product markets, and the competitive pressures these markets exert
are presumed to produce the set of contracts that maximizes the firm's value. As a normative matter the contractarian view implies that corporate law should not impose mandatory terms, which
would preclude contracting on a given topic, unless such terms are justified by identifiable flaws in
the bargaining process.363 The presumption of efficient corporate contracting is held with varying
strength by various commentators. There is some disagreement regarding what identifiable breakdowns in bargaining warrant mandatory terms, and whether the connections among participants in
the corporate enterprise may be deemed "contracts" in a meaningful legal sense.364 Setting the latter
361.
See Ronald Coase, The Nature of the Firm, in The Firm, The Market, And the Law 33-55 (1988)
(discussing the question of when market transactions are replaced by intra-firm transactions); Ian Ayres, Making A
Difference: The Contractual Contributions of Easterbrook and Fischel, 59 U. Chi. L. Rev. 1391, 1395 (1992)
(noting Coasean foundations of contractarian theory).
362.
The standard summary of this position is Easterbrook & Fischel, supra note 352, at 12-15. The term
“nexus of contracts” may be traced to Michael C. Jensen & William H. Meckling, Theory of the Firm: Managerial
Behavior, Agency Costs, And Ownership Structure, 3 J. Fin. Econ. 305, 310-11 (1976). The nexus concept itself is
commonly traced to Armen A. Alchian & Harold Demsetz, Production, Information Costs, and Economic
Organization, 62 Am. Econ. Rev. 777, 777-78 (1972).
363.
E.g., Easterbrook & Fischel, supra note 362, at 15; Klausner, supra note 5, at 760.
[Vol.
YEAR]
SHORT TITLE
127
debate aside for present purposes, we seek to determine whether network theory identifies a systemic tendency toward suboptimal contracting.
Klausner rightly notes that the contractarian paradigm assumes that "the value of a corporate
contract term is unrelated to the number of firms that adopt the term."365 As Klausner argues, to the
degree a firm's choice of governance of contractual terms is based upon the choices of other firms,
"market forces cannot be relied upon to promote socially optimal corporate contracts."366 It would
not necessarily follow, however, that a shift to some alternative presumption, such as adoption of
mandatory rules or different types of contractual defaults, would be warranted. Klausner quite appropriately refrains from making this analytical leap. Such a decision would depend not only on
the difficulties with the contractarian paradigm, but also on the viability of judicial or governmental
initiatives designed to remedy any systematically suboptimal outcomes.
2. Potential Challenges From Network Theory
Klausner posits several reasons why the value of a corporate contract term might increase with
the number of firms adopting the term. One group of these reasons, which he calls "interpretive
effects," is based on the notion that more frequent adoption of a given term will increase the probability that the term will be subject to future interpretation by courts. That judicial interpretation in
turn will presumably lead to greater clarity in potentially ambiguous terms.367 Klausner distin-
364.
On this point see Melvin Aron Eisenberg, Contractarianism Without Contracts: A Response to
Professor McChesny, 90 Colum. L. Rev. 1321 (1990) and Fred S. McChesney, Contractarianism Without
Contracts? Yet Another Critique of Eisenberg, 90 Colum. L. Rev. 1332 (1990).
365.
Klausner, supra note 5, at 761.
366.
Id. at 759.
367.
Id. at 776. Gordon makes this point as well. See Gordon, supra note 353, at 1566 n.56.
++128
CALIFORNIA LAW REVIEW
guishes interpretive effects from the inherent benefits of a term, comprised of its linguistic clarity
and all practices and precedents that exist at the time a firm is required to make a choice regarding
the term.368
In addition to the idea of risk reduction through enhanced clarity of contract terms, Klausner
applies his network effects analogy to certain economies associated with lawyers, accountants, and
investment bankers.369 Klausner rightly notes that such firms lower transaction costs in capital
markets by developing contract terms, norms of behavior, and similar structures to facilitate transactions even where the parties themselves have little or no experience. To the extent any of these
structures becomes a standard (such as the Generally Accepted Accounting Principles), there will
368.
As Klausner states:
At the time a firm adopts a contract term, past precedents are not network benefits. They are part of the term’s
currently understood meaning and, in network externality vocabulary, are therefore elements of its inherent benefit.
Klausner, supra note 5, at 776 n.61.
As almost a reciprocal of this point, Klausner argues that the more widely a given term is adopted, the more likely it
will be to give rise to accepted standards of conduct within the business community that may provide further clarity
with respect to implementation of the term. Business professionals and business lawyers take a considerable degree
of comfort in following procedures used by other prestigious firms, which in turn may have originated with, or at
least received the approbation of, the courts. Thus, as Klausner rightly notes, if courts are willing to accept an
investment bank’s fairness opinion at face value as evidence that a board of directors properly discharged its duty of
care in evaluating a control proposal, then such opinions will be routinely recommended by corporate lawyers; if
nothing else, the recommendation will show the lawyer’s own diligence and provide a record that can be defended in
court by citation to favorable precedent. See id. at 781-82; Claire A. Hill, Fool Me Twice, Shame on Me, Or How
Corporate Lawyers Learn from Experience (working paper 1997).
369.
See id. at 782-84.
[Vol.
YEAR]
SHORT TITLE
129
be some inertia weighing against either the adoption of any alternative standard or any deviation
from the standard through adoption of a term customized to fit the needs of a particular firm.370
Such deviation would require a firm to forgo the learning economies derived from its intermediaries' experience with standard terms. Klausner analogizes such economies to the production of
goods compatible with a network good.371 As a related but distinct point, Klausner identifies
learning economies among investors as "marketing network effects." To the extent investors have
become familiar with standard terms, Klausner notes that issuers will be reluctant to depart from
such terms out of concern that investors would not price the nonstandard term properly, would fear
that other investors would not do so (creating liquidity problems in secondary markets), or would
discount the terms to compensate for the cost of analyzing their value.372
These concepts must be treated with care. Leaving aside corporate charters, and focusing on
corporate contracts to which meaningful consent presumably has been given, such as a bond indenture, contracts are fundamentally artifacts (hopefully) embodying the understanding of the contracting parties. Unlike microprocessors or software, contracts do not themselves process data;
unlike telephone networks they do not themselves convey communications. The value of contracts
rests ultimately on human understanding and interpretation. Accordingly, the processes of interpretation and applied understanding will always be at work when a contractual term is at issue.
Klausner does an estimable job of parsing different elements of interpretation and understanding
and exploring the manner in which widespread adoption of similar terms affects these elements.
370.
The flip side of this, of course, is that information presented in GAAP form is more valuable to those
familiar with GAAP and what it represents than is information presented in another form.
371.
See Klausner, supra note 5, at 782 n.82
372.
See id. at 785.
++130
CALIFORNIA LAW REVIEW
Because these elements ultimately converge in an act or acts of interpretation bringing the contract
to life, however, the task of attributing causal significance to a network effect, and of prescribing
corresponding actions, is more complicated than would be the case for goods that require relatively
less human intervention to provide value to a user.
This point raises a related concern: the networks to which Klausner refers are comprised of
contractual and governance terms as such. Each of the effects Klausner discusses should be modeled as a service or input compatible with the contract terms—the "network good.” Because interpretive acts are necessary to derive value from contracts, and the effects Klausner identifies are all
elements of "interpretation," loosely defined, the distinction between the service and the good may
appear artificial, and is certainly less clear (in terms of value added) than the distinction between a
microprocessor and an operating system. Both the collective act of contract creation and the necessity of interpretation make attributing a particular consequence to a particular "network effect" difficult, thereby compounding the problem of determining whether and to what extent such effects
either undermine contractarian assumptions or suggest modifications to corporate law.
Klausner derives from these concepts a model describing possible suboptimal contracting.
Network theory itself supports any of four different outcomes: where standardization is desirable,
the optimal standard may be adopted, a suboptimal standard may be adopted, or the optimal standard is adopted but become suboptimal over time for reasons of inertia and collective action problems.
Alternatively, where standardization is undesirable, a standard may be adopted or
maintained, by definition a suboptimal outcome, or no standard may emerge, by definition an optimal outcome.373
373
Farell & Saloner, supra note 8.
[Vol.
YEAR]
SHORT TITLE
131
In Klausner’s model, where firms place the same value on terms (standardization is desirable)
firms may nevertheless adopt suboptimal terms (either one suboptimal standard or multiple different terms) that have high inherent value relative to the present value (discounted for the risk that
other firms will not adopt it) of an alternative term that would have greater value if it became a
standard through adoptions over time. Conversely, where firms place different values on a term,
they may nevertheless adopt a standard term that differs from the firm-specific optimal term simply
to obtain the benefits of the standard.374 Consistent with this array of possible outcomes, Klausner
is careful to note that the implications of possible network effects for the law are indeterminate.
In light of the analytical difficulties resulting from such indeterminacy, Klausner's recommendations for change in corporate law are appropriately exploratory in nature and scope: he calls for
further empirical research and suggests that corporate law might profitably gravitate away from a
paradigm of relatively specific default rules, designed to mimic the outcome of a hypothetical bargaining process, and instead move toward a standard-setting paradigm offering menus of relatively
tailored default rules.375 He also suggests that network theory may help explain Delaware's dominance in the market for corporate charters.376
Before analyzing these arguments in detail, we wish to make two points about Klausner's use
of the network analogy in corporate theory and his later work with Marcel Kahan on event-risk
bond covenants. First, we applaud the general nature of the inquiry. Without wishing to engage
even briefly in the debate over what aspects of economic theory properly fall under such rubrics as
"transaction cost economics" versus "neoclassical economics" or the "new institutional econom-
374
See Klausner, supra note 5, at 813.
375.
See id. at 839.
376.
See id. at 842.
++132
CALIFORNIA LAW REVIEW
ics," 377 we work from the premise that the Coasean origins of contractarian corporate theory dictate
that the theory should be focused on the costs of forming and altering corporate contracts—be they
governance terms or debt issues.378 Klausner and Kahan rightly recognize that their exploration of
possible influences of network effects on corporate contracting behavior is a step in that direction.
They seek
to crack open further the black box of the corporation by enhancing the understanding of why corporate contracts look the way they look, by critically examining
the extent to which atomistic corporate contracting can be expected to produce a socially optimal degree of standardization, and by investigating institutional factors
that may overcome the potentially suboptimal influences of atomistic corporate contracting.379
The exploration of network theory in corporate contracting is thus precisely the sort of work
that is required to test and provide substance to contractarian theory. Though, as we explain below, we would attribute less of the economic behavior they examine to network effects than to
377.
For more on this debate, see Richard A. Posner, Overcoming Law 426-43 (1995).
378.
As Oliver Williamson has pointed out, Coase himself did not dictate the level of specificity at which
transaction cost analysis should take place. See Williamson, supra note 18, at 230. But Coase’s general approach
and call for empirical research into corporate contracts suggest agreement with the concept of designating transactions
themselves as the appropriate unit of analysis. See Ronald H. Coase, Essays On Economics And Economists 13
(1994).
379.
Kahan & Klausner, Standardization, supra note 61, at 716.
[Vol.
YEAR]
SHORT TITLE
133
other related economic influences, we wholeheartedly agree with the need to examine more closely
the actual contracting behavior of firms and the economic forces that influence such behavior.380
Second, it is important to distinguish between the economic influences Kahan and Klausner
group under the heading of "learning benefits" and those they analogize to network benefits. The
difference is temporal. Learning benefits refer to the value imbued in a given term through previous experience. Such benefits include simple drafting efficiencies,381 reduced uncertainty due to
previous judicial interpretation, general familiarity with a term and, possibly, development of behavioral norms within the financial community.382 As noted above, for purposes of network theory, Klausner categorizes these as inherent rather than network benefits, reserving the latter term
for those benefits accruing through contemporaneous use of a term by other firms.383
Particularly with respect to interpretive benefits, this distinction is extraordinarily fine and may
be untenable. One may price an asset that has not generated past earnings solely by predicting future income streams and discounting for risk; one would be unlikely to replicate that analysis for an
asset that had generated income in the past. A more likely approach (absent a material change in
380.
Or, as Oliver Williamson has described the endeavor of operationalizing transaction cost theory, the
task entails “(1) identifying the critical ways in which transactions differ and explicating their economic significance;
(2) identifying the key features that distinguish alternative governance structures, and developing the basic tradeoffs
in moving from one to another; (3) working out the logic of the discriminating match, whereby transactions are
aligned with governance structures in a way that economizes on transaction costs; and (4) doing empirical studies
that bear on the predictions that are derived from these operationalization efforts.” Williamson, supra note 18, at 230.
381.
Though the cost of reproducing terms in the computer age is low, and the incremental efficiencies of
duplicating prior forms are therefore slight.
382.
See Kahan & Klausner, Standardization, supra note 61, at 719-24.
++134
CALIFORNIA LAW REVIEW
circumstances) would incorporate prior performance in the analysis. Similarly, as a practical matter, when called upon to decide whether to adopt or adhere to a given term, the view of lawyers
and bankers with respect to construction of the term will be influenced as much by past experience
(inherent benefits) as by future expectations (network benefits). And future expectations are themselves in significant part a product of past experience. Given that a relatively high ratio of network
to inherent benefits is what distinguishes network goods from other goods, the difficulty of drawing this distinction renders analysis of interpretive effects highly problematic.384 Even with respect
to service and marketing effects, network effects will likely be closely related to the historical development of a market;385 it therefore will be very difficult to evaluate the relative importance of inherent and network benefits in a given case.
While Klausner's model and analysis rigorously explore the potential for using network theory to inform corporate legal analysis, our more general analysis of network theory provides a
slightly different perspective and a correspondingly different emphasis on the importance of network theory for either the contractarian paradigm or existing legal rules. Based on our more general analysis, and subject as always to further empirical work in the field, we are skeptical that the
evidence Klausner (and Klausner and Kahan) find in support of their analogy in fact represents
evidence of network effects, and in particular of the claim that network effects might be the cause
of the potentially suboptimal circumstances they identify. We accordingly assign less weight to the
degree to which network theory either weakens the presumptions of the contractarian paradigm or
warrants changes in existing legal rules. In brief, as explained in the following sections, we be-
383.
See id. at 724-25.
384.
We discuss this point in greater detail in connection with section F.6, infra.
385.
See supra notes 49-51 and accompanying text (noting importance of history in network markets).
[Vol.
YEAR]
SHORT TITLE
135
lieve that network effects in corporate law will be difficult to identify as such, will be weak where
they can be found, and will likely be subject to amelioration through market forces.
3. Interpretive Effects
The interpretive network effects Klausner identifies are grounded in the economics of information and therefore apply in those circumstances in which information and information costs play
an important role in market structure—which is to say they apply in virtually all circumstances involving transactions.386 He characterizes these effects as a combination of the inherent imprecision
of language and the principle that imprecision, here another word for risk, tends to reduce the value
of any contract.387 In short, uncertainty is costly, and if widespread use of a contract term can reduce uncertainty then the term's value will increase to a level beyond its intrinsic value simply because it was widely adopted. If Klausner has identified systemic, network-based flaws in actual
contracting behavior, the relevance of his insights will be widespread if not ubiquitous in cases involving commonly-used contracts.388
In terms of network theory, the value of a contract term is comprised of its inherent benefits—
the innate clarity of the term and the value of past experience and judicial precedent—plus the pres-
386.
Cf. James Boyle, Shamans, Software and Spleens: Law and the Construction of the Information
Society 41-46 (1996) (noting the pervasiveness of information cost limits in modern economics). But see Lemley,
supra note 204, at 890-94 (contesting Boyle’s conclusions on this score).
387.
See Gordon, supra note 353, at 1566 (noting that imprecision of nonstandard corporate charter terms
will create costs due to risk but arguing that such costs are unproblematic from a contractarian perspective because
they will be internalized by the firm adopting the nonstandard term).
388.
This indeed is the thesis of Klausner’s further exposition of his arguments with Professor Kahan,
which sets out to discuss “the economics of boilerplate” as such. See Kahan & Klausner, supra note 61.
++136
CALIFORNIA LAW REVIEW
ent value of the potential for greater clarity in the future through litigation, which value increases as
more firms adopt the term, and which comprises the network benefit.389 Klausner hypothesizes
that firms shifting away from established standards will incur potentially material opportunity costs
in the amount of the present value of increased clarity through future litigation. By hypothesis that
value will be much lower for nonstandard terms than for standard terms.390
As a practical matter, we do not see how to parse a firm's valuation of the cost of losing inherent benefits (innate clarity plus precedents and behavioral norms) and the firm's valuation of the
opportunity cost of possible future clarifications. To the extent such parsing could be done, however, we tend to doubt that the costs of contracting around default rules or standardized contract
terms, including both the costs to negotiate and memorialize the nonstandard term and the loss of
future interpretive benefits, will be very high relative to the inherent benefits of any given term at
any given point in time. At a minimum, we would expect firms to place more weight on existing
389.
And, to be fair, less the present value of the probability that widespread adoption will reduce clarity
through the publication of inconsistent opinions.
390.
This view of course implies that we should be concerned with the structure of default rules—whether
they should be “tailored” (drafted with an eye toward a specific firm or subset of firms with a particular salient
characteristic) or “off-the-rack” (drafted to mimic the terms a majority or plurality of firms would agree upon if they
bargained over a term). Klausner argues that “[t]he presence of network externalities in corporate contract terms
militates in favor of tailoring to some degree” because even if an off-the-rack default “is value maximizing for a
majority or plurality of firms, its ability to promote the formation of a contractual network can nonetheless lead to a
socially suboptimal contract equilibrium.” Klausner, supra note 5, at 832. There is much to explore in this idea,
but we here simply note the issue and leave the exploration for another day.
[Vol.
YEAR]
SHORT TITLE
137
precedents than the uncertain possibility of future judicial clarification.391 We consequently suspect
that interpretive network effects associated with either default or standardized terms will be relatively weak, i.e., unlikely to represent a systemic flaw in the presumption of independent contracting.392
Assuming for the moment, however, that one can isolate inherent and network benefits with
respect to a given term, the key variable with interpretive effects is the clarity of a term's meaning.
The value of relatively specific terms—such as a cumulative voting default, the number of shares
authorized, permissible debt ratios, or the proportion of votes required to approve a merger—is
inherent, a point Klausner rightly recognizes.393 It follows that even if firms exhibit a high degree
of standardization with respect to specific terms, such standardization does not necessarily (or even
likely) imply the presence of network effects. Relatively specific provisions will by definition have
391.
If this intuition is correct, the concept of interpretive network benefits requires a reconceptualization of
network theory as well as corporate law, for it defines as a network good contract terms for which inherent benefits
can be expected to be systematically higher than network benefits.
392.
We can dispense with inherent switching costs relatively quickly. In contrast to cases involving large
capital investment in a particular network technology, the actual expenses incurred in adopting a non-standard term
are not likely to be material. A contract term is not a computer program: discarding an old one in favor of a new one
does not require the investment of any particular resources, or the conversion of old data. And while contract terms
might theoretically be protected by copyright or even patent law, in practice lawyers freely copy contractual
innovations from other lawyers without paying for them.
393.
For example, if contracts could be written and interpreted by computers in binary code, the terms
themselves would not be subject to network effects; nor would contract terms governed by mathematics, such as a
bond covenant preventing an issuer from assuming debt beyond a certain ratio relative to its assets or cash flow.
Klausner’s model incorporates this prediction as well. See Klausner, supra note 5, at 763-65.
++138
CALIFORNIA LAW REVIEW
relatively little value to gain through possible clarification by litigation or other forms of interpretation. This holds true both because the process of interpretation would have little to add to the clarity of such terms and because such terms, being clear, would be relatively unlikely to give rise to
disputes requiring interpretation. The opportunity cost of departing from a specific standard would
therefore be quite low, as would the risk of a socially suboptimal equilibrium with respect to such
terms.
The case for interpretive effects therefore rests on the value of possible future clarification of
less precise terms—such as fiduciary duties, or the rule that incumbent management may oppose
unwanted tender offers so long as the measures adopted are "reasonable in relation to the threat
posed."394 Such rules are more likely to give rise to litigation precisely because they are vague and
therefore could potentially support multiple courses of action. The prospects for future litigation to
clarify such terms will be stronger but, as a practical matter, we doubt whether such prospects pose
a real challenge to the presumptions of the contractarian paradigm or warrant changes in corporate
governance statutes.
While we do not take issue with the general proposition that interpretation through litigation
may add some value to open-ended contract terms by increasing their clarity, there are fairly strict
limits to the value of such effects, particularly when one counts the existing body of precedent as
an inherent benefit. Again assuming that one can draw this distinction, we doubt that the marginal
gains in clarity that might be obtained through future litigation are high enough to create benefits
sufficient to lock firms into any particular term, particularly when compared to whatever incentives
the firm has based on the existing body of precedent and experience. This is particularly true when
the inherent meaning of the potential terms differs: no firm will choose a rule that will disadvan-
394
Unocal Corp. v. Mesa Petroleum, 493 A.2d 946, 949 (Del. 1995).
[Vol.
YEAR]
SHORT TITLE
139
tage it because that rule is likely to be clearer. In other words, given relatively high inherent benefits, as a practical matter there will be no way to tell whether a firm adopts or retains a given term
because of any identifiable network benefits, and thus no way to conclude that network benefits are
the cause of any particular equilibrium.
There is in any event no a priori reason to believe that a significant portion of the disputes over
the meaning of a term will in fact be resolved through litigation and, more specifically, through reported appellate decisions. The high cost of litigation, the reputational cost of being known as
someone who sues their business partners, the probability in the relatively small world of corporate
finance that two parties will desire to do business in the future, and the availability of less expensive means of alternative dispute resolution, that do not produce precedential decisions, all combine
to make the litigation model one of last resort in many instances.395 In light of such disincentives to
litigate disputes through appellate courts, even a nontrivial increase in the probability that a term
will be litigated may not increase the likelihood of enhanced clarity very much.396
In addition, even reported decisions may not add much to the inherent value of open-ended
terms. In at least some instances, open-ended governance terms—particularly governance defaults
for large, publicly-traded firms with multiple traunches of debt (and on occasion different classes
of stock)—will resemble the antitrust statutes: majestically general language overlaying a byzantine
395.
For a detailed elaboration of this point, see Edward L. Rubin, The Nonjudicial Life of Contract:
Beyond the Shadow of the Law, 90 Nw. U. L. Rev. 107, 118-22 (1995).
396.
As Klausner points out, however, to the extent material gains in clarity may be obtained through
litigation, the small probability that any given dispute will lead to a reported decision supports the view that any
interpretive effects will accrue throughout the full range of adoptions of that term. For interpretive effects, in other
words, any network benefits would not be inframarginal, as any benefits associated with legal service and marketing
effects almost certainly are. See infra text accompanying notes 412-454.
++140
CALIFORNIA LAW REVIEW
mixture of economic relationships which poorly understood by the parties to the various relationships, to say nothing of jurists called upon to make sense of multiple competing positions.397 Such
general language has not historically been a hospitable medium for the development of clear legal
meaning, regardless of the number of times courts interpret a term. Indeed, where courts interpret
general terms (such as "reasonable" or "threat") to achieve what they perceive to be a just result in a
particular case,398 there may well be an inverse relationship between the probability of obtaining a
reported opinion addressing a given term and any expectation that the term will be further clarified.399
397.
And this assumes all the parties and the court succeed in identifying and arguing the relevant issues.
On occasion, the central issue in a case is missed or waived by one of the parties, producing skewed decisions that
distort the law and impede development of the coherence necessary to the task of adapting network theory (or any
other theory, for that matter) to the law. See McGowan, supra note 10, at 787 (noting that counsel on appeal in
Aspen Skiing failed to preserve the question whether the defendant was a monopolist, which it almost certainly was
not).
398.
Which is not to say that a court seeking such a result would hand down inefficient decisions. As Ian
Ayres has noted, “[c]ourts can promote efficiency by imposing contingent contractual obligations ex post that
corporate stakeholders are in practice unable to contract for ex ante.” Ayres, supra note 361, at 1404.
399.
See, e.g., Posner, supra note 22, at 6 (discussing the failure of antitrust law to achieve much clarity in
evaluating complex economic arguments). Klausner of course does not contend that adoption of a standard will lead
to “clarity” in any abstract, normative sense; his argument advances the more limited and more realistic point that
widespread adoption will yield marginal increases in clarity. Again, in this instance our view reflects a differing
assessment of the likely magnitude of such effects rather than with the logic underlying Klausner’s argument as
such.
[Vol.
YEAR]
SHORT TITLE
141
Broad concepts such as the duty of care encompass so many different situations that it will
often be difficult simply to identify what issues the courts believe they are addressing. Should one
view Smith v. Van Gorkom, 400 for example, as a standard duty of care case or as a subspecies pertaining to control transactions?401 Would Time's asserted right to preserve corporate policy fare as
well for a tire manufacturer as it did when imbued with a heavy dollop of First Amendment rhetoric?402 Tests under which such concepts are evaluated tend to be accordingly flexible. The Delaware Supreme Court in Time Warner, for example, praised the Unocal test403 precisely for its
elasticity: "The usefulness of Unocal as an analytical tool is precisely its flexibility in the face of a
variety of fact scenarios. Unocal is not intended as an abstract standard; neither is it a structured
and mechanistic procedure of appraisal."404
400.
488 A.2d 858 (Del. 1985) (also referred to as the “Trans Union” case).
401.
See Ronald J. Gilson & Bernard S. Black, The Law And Finance of Corporate Acquisitions 1055
(noting that “part of the puzzle is whether Trans Union really is a business judgment case at all”); Jon Macey &
Geofrey Miller, Trans Union Reconsidered, 98 Yale L.J. 127, 128 (1988).
402.
571 A.2d 1140 (Del. 1990).
403
See Unocal Corp. v. Mesa Petroleum, 493 A.2d at 946.
404.
571 A.2d at 1153. Klausner acknowledges that a fair degree of litigation in the Delaware courts has
brought little in the way of clarity to the rules governing permissible defenses to takeover attempts opposed by
incumbent management. Klausner, supra note 5, at 762 n.12. As Unocal shows, fact-specificity may prevail even
within a genre of cases. Taking Chancellor Allen’s analysis in Interco as an example, see Capital City Associates
Limited Partnership v. Interco, Inc., 551 A.2d 787 (Del Ct. Chanc. 1988), the question whether an offer poses a
threat may (and arguably should, notwithstanding the Delaware supreme court’s chastising words in Paramount I)
rest on the economic characteristics of the offer: an all cash/all shares offer may be considered non-threatening per
se, though that is at least arguably not the present state of the law. A share exchange offer may or may not be the
++142
CALIFORNIA LAW REVIEW
These limitations are reflected in the principal example of interpretive effects Klausner cites for
the proposition that network effects cause suboptimal contracting: the "apparently widespread use
of ‘plain vanilla’ charters, and hence the wholesale adoption of default rules, as opposed to customized corporate charters."405 Klausner is careful to note that this example is to some degree
speculative, based on the intuition that charters merely embracing the law of a given state are more
widely adopted than the heterogeneity among firms suggests that they should be.406 As he puts it,
interpretive network effects analysis "suggests that there may be terms for which some degree of
customization is clearly permissible and potentially value enhancing, and yet firms will still decline
to customize those terms."407
It is not clear why interpretive network effects carry much explanatory power with respect to
the adoption of "plain vanilla" charters, or the more general failure of firms to customize their
charters to embrace alternatives to state-law default rules, particularly when the body of precedent
and business norms relating to such terms is an inherent benefit of those terms at the time of adoption. The idea that a firm adopting a charter would value the prospect of future clarification of ex-
equivalent of an all-cash offer; an offer incorporating some mixture of cash, stock, and debt similarly may or may
not be the equivalent of an all-cash offer. The number of variables necessary to the determination whether a threat
exists increases in the latter two cases; with such increases comes the increased possibility of distinguishing
precedent, and increased complexity in determining whether the clarity of a governance term has been enhanced,
diminished, or unaffected by a given decision.
405.
Klausner, supra note 5, at 822. Klausner also presumes that some interpretive network effects existed
with respect to certain indenture terms, though his discussion of the model debenture indenture provisions pertains
more to legal service effects than to interpretive effects. Id. at 817-18.
406.
See id. at 822.
[Vol.
YEAR]
SHORT TITLE
143
isting rules more than the extant precedent seems counterintuitive.408 On the other hand, very little
about the operation of interpretive effects is clear, as shown by a more specific example Klausner
uses to illustrate suboptimal contracting. He suggests that the iteration of the Unocal test adopted
by the American Law Institute's Principles of Corporate Governance may not have been adopted
by many firms (which presumably retained "plain vanilla" charters) because the ALI's version is
"replete with open-ended formulations" and thus a large portion of the provision's value lies in future interpretations, which are in turn a function of the number of firms adopting the test.409
As Klausner notes, however, the ALI's test is simultaneously "broadly consistent with the
Unocal test" and clearer than Unocal in significant ways.410 If relative clarity (in present value
terms) is in fact the desideratum of firms choosing governance provisions, one would not expect
interpretive effects to prevent firms from adopting the ALI's test. After all, both that test and Unocal offer future benefits of clarity through litigation, while the ALI's version offers some marginal
407.
Id. at 822-23.
408.
In addition, to the degree that firms view charters as a whole, rather than individually, they could quite
rationally conclude that the value of the charter was on balance inherent (more in specific than ambiguous terms) or
that the cost of sorting the different types of terms and tailoring the latter exceeded any benefits. In either event,
firms would not be swayed in their decision by relatively lower-value interpretive benefits. To be fair, however,
even if there are more specific terms in a charter than general terms, the latter are likely to be more important in a
qualitative sense.
409.
Klausner, supra note 5, at 824.
410.
Id. at 823-24 (noting that the ALI test “dispenses with the vague and inapposite ‘corporate policy and
effectiveness’ language” of the Unocal test and “explicitly states that the directors’ action must be in the best
interests of the ‘corporation and the shareholders’ and it defines that term to mean the enhancement of profitability
over the long term”).
++144
CALIFORNIA LAW REVIEW
advantage in clarity over the status quo. The ALI test effectively incorporates much of the existing
precedent interpreting Unocal and, where it does not, its departures tend toward relative clarity.
Given the broad consistency of the ALI and Unocal tests, the ALI test would appear to have a
greater inherent value (incorporation of existing precedent and relatively greater innate clarity where
existing precedent was eschewed) than the original Unocal test. As Professor Klausner notes, the
ALI test is also open-ended in material ways, and thus should accrue interpretive benefits at least
roughly equivalent to those expected for Unocal itself.411 Yet firms have not shifted. The answer
may be a simple agency cost story: managers may prefer the freedom to maneuver offered by the
ambiguous aspects of the Unocal test and, having control of such matters, refuse to shift. Or there
may be another explanation. Whatever the reason, interpretive network benefits would appear to
be an unlikely causal candidate. Indeed, if anything such effects would suggest the opposite result.
4. Service Providers
The network analogy is further complicated by the operation of what Klausner calls servicerelated network effects, which are in essence inputs to the operation of the governance or debt
terms at which Klausner's (and Klausner and Kahan's) arguments are directed.412 In operational
terms, these effects involve the business behavior of law firms and investment banks who, often
unlike the corporations for whom they work, are repeat players in corporate financial transactions.
411.
If Unocal accrued greater interpretive benefits, then the failure of firms to switch to the ALI’s test
would be attributable to network benefits, but in a sense opposed to the manner in which Klausner analyzes the case.
412.
These services should be considered inputs regardless of whether they are obtained from retained
advisers or from employees of a firm. Whether a given firm will bring such functions in-house is presumptively a
function of its cost structure, but the service rendered is an input to the network in either case.
[Vol.
YEAR]
SHORT TITLE
145
In Klausner's view economies associated with scale and learning effects may enable such firms to
provide more efficient services for widely-adopted terms than for customized terms. Firms adopting the standard will enjoy the benefits of such efficiencies while firms adopting customized terms
may encounter relatively greater costs.413
Our first difficulty with this view is that such economies do not reflect very strong network
effects and likely will not be the cause of any suboptimal contracts. As discussed above, legal and
investment banking services are compatible with "networks" of terms; Klausner thus rightly conceives of service effects as economies of scale in an input or complementary good, rather than as
direct network effects. 414 In other words, multiple adoptions of a term may provide a sufficiently
large customer base for service providers to realize such learning or scale economies as are available. Legal and financial service economies of scale are thus indirect "market-mediated" effects, to
413.
See Klausner, supra note 5, at 782. In addition, though not a part of Klausner’s network analogy, to
the extent lawyers and bankers effectively dictate the governance or debt terms a firm will adopt, the lawyers’ and
bankers’ own interests will tend to drive firms to adopt standard terms. It is possible that lawyers and bankers would
enforce a suboptimal standard because other lawyers and bankers did so, even if the standard had few or no network
benefits to clients actually adopting governance terms. See Kahan & Klausner, Path Dependence, supra note 353.
Hence, “network benefits” to lawyers and bankers would appear as agency costs to firms.
Though analytically
distinct from such value as may inhere in governance terms themselves, the incentives of lawyers and bankers could
in theory reinforce the tendency toward standardization implied by Klausner’s interpretation of network theory.
414.
This distinction becomes even harder to draw with respect to lawyers and bankers because, as a
practical matter, it is likely more often than not the case that lawyers and/or investment bankers are the drafters of
the terms, the entities that choose which terms will be adopted and, should a dispute arise, the entities that will
interpret the terms for a firm.
++146
CALIFORNIA LAW REVIEW
borrow from Farrell & Saloner,415 and are materially weaker than direct network effects, or even
indirect network effects in the presence of actual interoperability, as is the case with computer
software. 416
Regardless of the abstract strength of such effects, however, we see very little reason to believe that service effects will entice firms to deviate from socially optimal contracts. Nor do such
effects warrant an alteration in existing law. Klausner recognizes that legal service effects are
likely to peak earlier than interpretive effects;417 we agree with this point but would extend it further. Both legal and investment banking service effects would likely peak at a level far below the
extent of the market as a whole, resembling what Liebowitz and Margolis dub "inframarginal network effects." 418
In other words, one would expect the learning curve of lawyers and bankers with respect to
model indenture terms, for example, to drop fairly rapidly and to become immaterial well short of
the extent of such terms in the market as a whole. Beyond a fairly modest level, therefore, additional issuers adopting such terms are unlikely to create incremental gains in legal or financial
service efficiencies and are thus unlikely to confer benefits on other firms using those terms. In
addition, both by training and experience lawyers and bankers should be capable of interpreting a
415.
See Farrell & Saloner, supra note 8, at 70.
416.
As noted, Klausner rightly analogizes service effects to the production of compatible goods, and thus
rightly distinguishes his analogy from direct network effects. We believe he overstates the strength of such effects,
however, by comparing them to software/hardware interfaces, which are subject to interoperability effects that are
material to the strength of a network but are absent, except at the level of pure (nonproprietary) language, in legal
and financial services.
417.
See Klausner, Networks, supra note 5, at 784 n.87.
[Vol.
YEAR]
SHORT TITLE
147
wide array of terms. The marginal difference in efficiency between advice on standard and nonstandard terms is therefore unlikely to be very large regardless of learning effects. In any event, if
service effects are in fact inframarginal, they will likely be internalized through competition among
lawyers and bankers, so there is little reason to believe that the premises of the contractarian paradigm are materially weakened or that positive corporate law is in need of change.419
Indeed, competition among lawyers and bankers may ameliorate the lock-in problem that is the
prototypical inefficiency associated with network markets. Such professionals compete on a variety of grounds, and will seek every opportunity to distinguish themselves on grounds other than
price, including expertise. Thus, firms have an incentive to internalize the costs of familiarizing
themselves with a variety of terms.420 In addition, such firms often compete with one another by
418.
Liebowitz & Margolis, Uncommon Tragedy, supra note 5, at 140-41.
419.
This is particularly true with respect to terms with high inherent value. If such value arose from
innate clarity, then a professional would likely need little more than the education necessary to enter the profession
to deal adequately with the term. If such value arose through a body of precedent, one would expect lawyers in
particular, and bankers to a lesser extent, to familiarize themselves with the precedent as part of the normal
continuing education activities required to stay competitive in highly-competitive markets. There are many ways to
fail in a pitch for business; displaying ignorance of recent developments is near the top of the list.
420.
Cf. Hill, supra note 368, at 1 n.2 (noting that in practice forms are less standardized even within firms
than one might expect). One could of course argue that such behavior would reflect a different sort of externality—
excessive investment by firms in legal or financial education with respect to a suboptimally diverse universe of
terms. Given the material limits of interpretive effects, however, it is quite likely that a substantial portion of such
investments will be incurred even if standard open-ended terms are used (and such investment is unnecessary for
specific terms). Once one isolates the marginal expenditure to deal with different wording for what may be the same
concept, the effects at issue appear to be quite weak.
++148
CALIFORNIA LAW REVIEW
offering innovative services and products such as hostile takeover expertise,421 or poison pills,422 or
bonds with "poison put" provisions.423 Competition through innovation may internalize much of
the cost of innovation, to some degree overcoming collective action problems that might otherwise
block transitions among standards in other circumstances.424 Though the evidence is anecdotal,
experience among intermediaries over the past 20 years suggests that competitive forces will in fact
facilitate innovation.425
421.
See infra note 425 (noting first-mover advantages accruing to the Skadden and Wachtell firms through
earlier practice in hostile transactions).
422.
See Gilson & Black, supra note 401, at 740-41 (noting origin of poison pill with Wachtell firm and
quoting memorandum from Wachtell to clients describing its “share purchase rights plan”).
423.
See Kahan & Klausner, Antitakeover, supra note 353, at 738 (discussing poison puts).
424.
Kahan and Klausner recognize this point with respect to poison puts as well. See id.
425.
In this sense, competition among intermediaries in financial markets is the competitive equivalent of
software reverse-engineering or “bridge” programs written to facilitate transition among applications. The rise of
both law firms and investment bankers willing to compete by expanding the range of services offered and undertaking
aggressive maneuvers they previously eschewed provides some reasonably persuasive evidence that the profit motive
will be strong enough in enough intermediaries to overcome risk aversion and thereby facilitate innovation. A good
account of this dynamic may be found in Ron Chernow, The House of Morgan: An American Banking Dynasty And
The Rise of Modern Finance 599-600 (1990). As Chernow describes the process by which Morgan Stanley entered
the business of advising clients on hostile transactions, work previously scorned by old-line firms, an influential
partner “made a pitch for hostile takeovers as an irresistible trend that was fair to shareholders, if not always to
management. The argument of inevitability was probably the decisive one. As one partner recalls, ‘The debate was,
if we don’t do what our clients want, somebody else will.’” Chernow goes on to note the influence of law firms that
were formed by castoffs from white-shoe firms and those whom the major firms would not hire—in the 1940s and
1950s this wide range of unacceptable applicants included Jews. “At this juncture, Morgan Stanley made another
[Vol.
YEAR]
SHORT TITLE
149
We believe that Klausner’s own example—the terms of corporate indentures—is more consistent with these explanations than with network effects. The corporate indenture story centers on
the American Bar Foundation's Corporate Trust Indenture project, which was designed to standardize the language of certain indenture terms that were unlikely to vary in value relative to different firms ("boilerplate") but that had previously differed among firms.426 According to Klausner,
"prior to the Project's inception, bond indenture terms were suboptimally diverse. The complexity
of the terms and the procedures they could trigger created legal network externalities."427 By this,
Klausner means that "the optimal contractual equilibrium with respect to boilerplate terms was one
in which all firms adopted the same boilerplate. As a result of uncoordinated contracting decisions
unorthodox decision. Like Morgan Guaranty, the firm had long relied on the white-glove law firm of Davis, Polk,
and Wardwell, which had looked on takeover work as vulgar and avoided it. With Morgan Stanley partners terrified
of lawsuits ensuing from takeover work, they now wanted a tough, seasoned specialist. Greenhill [the influential
partner mentioned above] insisted on hiring the experienced Joe Flom, of Skadden, Arps, Slate Meagher &
F l o m . . . . [Flom had] pioneered in hostile takeovers in the 1950’s, when Skadden, Arps was still a humble,
four-man operation. For twenty years, he thrived on the scraps from law firms that were too haughty or too
dignified to conduct hostile raids. . . . When Flom was made a special counsel to Morgan Stanley, there were
stormy scenes with Davis, Polk partners, who were deeply offended by the decision.
Whatever its other
consequences, the trend in hostile takeovers democratized the New York legal world and provided an opening in Wall
Street for Jewish lawyers. Both Joe Flom and Marty Lipton of Wachtell, Lipton, Rosen & Katz profited from the
early refusal of old-line Wasp firms to sully their hands with takeovers.” Id.
426.
See, e.g., Churchill Rodgers, The Corporate Bond Indenture Project, 20 Bus. Law. 551 (1965).
427
Klausner, supra note 5, at 817-18.
++150
CALIFORNIA LAW REVIEW
of individual firms and their lawyers, however, suboptimal diversity existed."428 The suboptimality
here is a failure to standardize, and thus to capture available positive externalities.
It is hard to see why the state of bond indentures before the ABA model was drafted reflects
suboptimal contracting, particularly in light of the increasing returns to standardization implied by
network theory. By definition a network good is one that increases in value with the number of
users adopting it. Absent marginal transaction costs exceeding the marginal benefits to standardization, the potential for increased value should drive network markets toward standardization.429
Why then, if the value of corporate indentures would have increased through standardization, did
firms persist in using a suboptimally diverse array of terms before the American Bar Association
intervened? In what sense can we say that network externalities caused such suboptimal diversity?430 With respect to causation, the very essence of the network concept implies the contrary
conclusion: indenture terms remained suboptimally diverse despite (not because of) the possibility
of increased value through standardization.431 The idea of uniformity certainly was familiar to cor-
428.
Id.
429.
Recall the characterization adopted by Katz and Shapiro: “[i]n markets with network effects, there is
a natural tendency toward de facto standardization, which means everyone using the same system.” Katz & Shapiro,
Systems Competition, supra note 25, at 105.
430.
On the question of causation, Klausner intends the indenture example as one of the failure of firms to
realize potential efficiencies. The term “externality” thus refers to unrealized value as such rather than to the forces
that prevented firms from realizing that value by adopting a standard.
431.
Within the standard network theory account, one suboptimal outcome that might be caused by
network effects would be the failure of a network to form because of collective action problems: consumers would
rationally delay purchasing until the network matured, thus presenting a chicken-and-egg problem.
Unlike a
consumer contemplating which VCR to purchase, however, corporations wishing to issue debt simply have to make
[Vol.
YEAR]
SHORT TITLE
151
porate lawyers: in the wake of the Trust Indenture Act of 1939,432 qualified indentures were required to include a variety of statutorily-mandated terms, which by definition could not vary in
material ways among indentures.433
As we do not see the existence of model debenture terms, or the state of indentures before the
model terms were adopted, as supporting the view that network effects are likely to prevent firms
from maximizing firm value through contracting.434 To the contrary, while we recognize that some
value may have been added to indentures by the ABA's Model Debenture Indenture, we see the
slow adoption of this standard as evidence either that the benefits of standardization were slight, or
that the absence of efficient data reproduction and storage technology such as copiers or computers
simply prompted each law firm or underwriter to continue using its own forms (a sort of technology-driven path dependence). As we see it, if firms are locked in to suboptimally diverse terms for
reasons of path dependence, that state of affairs tends to weaken the case for network effects, not
support it.
a decision on the terms the debt will carry. Delay is not an option. One would therefore expect that, over time,
firms would tend toward standardization on their own if network effects were truly present in any material sense.
432
15 U.S.C.A. § 77aaa et seq.
433.
Indeed, in the course of seeking expert witnesses one of us (McGowan) spoke to a long-time indenture
lawyer and former SEC commissioner who recounted hours spent as a young lawyer respondent-reading (with his
wife and newborn child, no less) new indentures against old to ensure that the TIA would not be violated by
typographical error. Whatever hardships technology has inflicted on young lawyers in recent years, and there are
plenty, we at least have been spared that indignity.
434.
To the extent that network theory points out possible benefits from standardization not seen in the
marketplace, the question becomes what forces impede standardization; legal analysis in that case would have to
address those forces directly to determine what, if anything, should be done.
++152
CALIFORNIA LAW REVIEW
5. Marketing Effects
Klausner also discusses the rise and fall of event-risk or "poison put" bond covenants, a topic
that his later paper with Marcel Kaham explores in greater detail. These covenants provide the best
context for discussing what Klausner describes as “marketing network effects”—essentially learning economies of scale among investors.435 During the relatively brief period when they were
common, event-risk covenants provided additional security for bondholders against a diminution
of the value of their bonds due to a change in control of the issuer, such as occurred with the leveraged buyout of RJR-Nabisco.436 Klausner notes that such covenants were rapidly adopted in many
bond issues during 1988 and 1989 and then declined in use as takeover activity diminished in the
1990's. 437 For Klausner, this pattern is "consistent with the presence of network externalities in
these covenants. Both their rise and fall followed the bandwagon pattern associated with network
externalities. In addition, most details of the covenant quickly became standardized to a great extent, again indicating the presence of network externalities."438
We view the poison-put story as tending to contradict rather than support the network theory.
The bond indentures into which event risk covenants were introduced were presumably by and
large the progeny of the American Bar Foundation model debenture indenture, which Klausner
views as an instance of efficient standardization. If such indentures actually represented network
standards, one would expect some degree of inertia inhibiting departures from the standard via the
435.
See Kahan & Klausner, Standardization, supra note 61, at 726.
436.
See Kahan & Klausner, Antitakeover, supra note 353, at 934.
437.
Klausner, supra note 5, at 819.
438.
Id.
[Vol.
YEAR]
SHORT TITLE
153
introduction of new and untested provisions.439 The rapid rise of such covenants, which Kahan
and Klausner describe as being "fairly complex and differ[ing] significantly from other covenants," 440 thus requires some explanation.
The answer likely lies in the specificity of that portion of the covenants that departed from the
standard indenture. Kahan and Klausner believe that, though the risk of loss in bond value from
issuer control transactions was known before 1988 and some event-risk covenants existed, the
widespread attention and large decrease in bond values surrounding the RJR leveraged buyout
"may have increased the perceived risk of takeover-related losses and thereby enhanced the inherent value of the covenant."441 Bearing in mind that network goods are network goods precisely because their inherent value is low relative to their value as part of a network, an increase in the
inherent value of a term likely would increase the probability that it would be adopted but also
would tend to diminish the significance of any associated network benefits. In other words, high
inherent benefit was a sufficient reason for firms to adopt the terms, regardless of what other firms
did. Kahan and Klausner characterize the covenants as "relatively specific,"442 and we agree with
that characterization. 443 As noted above, the specific nature of such covenants supports the view
that the value of the terms was largely inherent.
439.
With respect to corporate contracts this is likely the only cost worth considering. Drafting costs
(including both counsel’s time and that of their word processing department) are most likely trivial relative to the
amount of any given debt issue.
440.
Kahan & Klausner, Standardization, supra note 61, at 746 n.72.
441.
Id. at 745.
442
See id. at 746 n.72.
443.
Kahan and Klausner do, however, envision that litigation could produce interpretive effects with
respect to the triggers of an event-risk covenant: “[i]ssues could have arisen, for example, related to the calculation of
++154
CALIFORNIA LAW REVIEW
Nevertheless, as the quotation from Klausner points out, network theory does predict that
once a new standard emerges it will be adopted very rapidly—that is the essence of the "tipping"
concern. Given the specificity of the terms, rapid adoption of a standard alone provides weak evidence in support of interpretive effects: high inherent value would be enough to attract firms to
adopt such terms. Suppose, however, that the RJR transaction did in fact awaken the market—or
more precisely the investment banking community—to the possibilities for a new (or at least revised) product that could be sold to both investors and management as protection against a common enemy.444 To us, this would seem to reflect a standard competitive dynamic and, perhaps,
bounded rationality on the part of investors in evaluating bonds prior to 1988. Kahan and Klaus-
the threshold shareholding or the measure of substantiality in an asset sale under the designated event provision.” Id.
at 746 n.72. We are skeptical that a sufficient number of disputes would arise under such provisions to yield
meaningful increments in the clarity of such provisions. The shareholding threshold (the percentage of stock that,
when acquired by a third party, would trigger a put) is a mathematical provision that should be susceptible of
adequate clarification at the drafting stage. While the sale of all or substantially all assets is indeed a vaguer term, it
is also a term included in the American Bar Foundation Model Debenture Indenture (§ 8-1) and therefore cannot be
considered to have any effects unique to event-risk covenants.
What value it had would be more properly
characterized as inherent, obtained through incorporation of terms that draw on precedent from model debentures and
from asset-sale provisions such as Del. Gen. Corp. L. § 271.
444.
There is some reason to be skeptical that this was the case. The senior partners in investment banks
and law firms in 1987 were the same individuals who cut their professional teeth on the conglomerate merger wave
of the 1960’s, and who had first-hand experience with hostile techniques and defenses going back at least that far.
Wall Street may indeed have a short memory—a nontrivial case could be made for limiting it to three months at
most—but the tactical considerations underlying the rise and fall of event risk bonds were not novel to those
involved.
[Vol.
YEAR]
SHORT TITLE
155
ner look at the event from a different perspective, arguing that press reports of the decline in RJR
bond values created an expectation that future debt issues would include event-risk covenants and
that "an expectation of widespread use may have increased confidence among early-adopting firms,
and early bond purchasers, that the covenant would be priced in the secondary market if bonds
with the covenant were sold several years later."445 They believe this expectation would have made
earlier investors more amenable to lower interest rates, thus lowering issuers' costs of capital.446
In terms of Klausner's model this argument rests primarily on marketing effects, which are in
essence demand-side learning economies. As with learning economies associated with service
providers, we would expect these economies to be inframarginal—only a relatively modest fraction
of all investors need familiarize themselves with such terms in order to provide liquidity in secondary trading. Past the point at which liquidity may be maintained, additional investors do not confer value on earlier investors by familiarizing themselves with the terms. Moreover, because the
unique aspects of event-risk covenants were relatively specific, there is no reason to expect learning costs to have been very high. Demand-side learning economies would therefore likely peak
very early indeed. This factor may also explain why the covenants fell out of use so quickly: if the
perceived rise in inherent value was triggered by investors' perceptions of the risk of highlyleveraged takeover activity, a fall in that risk would diminish the perceived inherent value of the
covenants. The high ratio of inherent to possible network value seems better able to explain both
the rise and fall of event-risk covenants than does a theory resting on network effects. Event-risk
445.
Kahan & Klausner, Standardization, supra note 61, at 746.
446.
See id.
++156
CALIFORNIA LAW REVIEW
bonds therefore provide weak evidence for a finding that network effects caused suboptimal contracts to be formed.447
We find further support for this view in Kahan and Klausner's contention that event-risk
covenants were affected by switching costs. Their findings are based on "intra-company consistency," meaning that once a firm had issued debt including event-risk covenants it was unlikely to
change its formulation regardless of innovations in such covenants elsewhere in the market: "once
a company adopted an event risk covenant, in essentially all cases it declined to adopt any subsequent improvements."448 Kahan and Klausner view this as "strong support for the hypothesis that
switching costs were present" and thus "indirect evidence for the presence of learning or network
benefits."449 While such behavior is consistent with an explanation based upon intra-firm learning
curves, and with the high inherent benefits of such terms, it is inconsistent with network theory,
which predicts that as standards become established firms will gravitate to them in order to obtain
447.
Kahan and Klausner note that the major bond rating agencies began to rate the protective qualities of
event risk covenants once significant numbers of the covenants began to appear, noting that “[t]he rating agencies
could presumably justify doing this only if a critical mass of bonds with these covenants were issued.” Kahan &
Klausner, Standardization, supra note 61, at 746 n.73. As with investors themselves, there is little reason to believe
that the scale necessary to amortize the agencies’ costs of familiarizing themselves with event-risk covenants was
sufficiently large to suggest the presence of an externality.
448.
Id. at 752.
449.
Id. Kahan and Klausner discount the possible explanation that each firm had adopted uniquely suitable
covenants on the ground that each firm retained its covenants in the face of materially improved alternatives. See id.
at n.83. We have no reason to disagree with this conclusion and would not find it surprising for a firm simply to
continue with prior issues it had successfully sold until the market would no longer buy them at an acceptable (if not
optimal) price.
[Vol.
YEAR]
SHORT TITLE
157
the benefits of standardization—again, this is the entire point of market "tipping."450 If the ratio of
inherent to network benefits is such that firms are content to adhere to terms once adopted, then
there is considerable reason to question whether a "standard" term in the market should be considered a standard for purposes of network theory.
We thus return to the problem of isolating the causal forces behind the observed behavior.
Kahan and Klausner characterize the effects in the alternative—as showing "the presence of learning or network benefits."451 Standard network theory suggests that the former effects are likely at
work here (which would be consistent with a finding of high inherent value), but that the latter are
not. 452 For these reasons, we disagree with Kahan and Klausner's conjecture that "[t]he presence
of such internal learning and network benefits may in turn be suggestive of similar (albeit lower)
external learning and network benefits."453 To the extent firms adhere to their own forms because
450.
Even from a marketing effects point of view, one would expect firms to prefer standard terms, with
which the secondary markets would by hypothesis be more familiar, than nonstandard terms regardless of intra-firm
forces. Indeed, the failure of early adopters of terms to adopt subsequent improvements by others tends to contradict
Klausner’s theory that the early adopters acted in order to take advantage of predicted network effects, since they did
not in fact follow those network effects when they actually appeared.
451
Kahan & Klausner, Standardization, supra note 61, at 752.
452.
Kahan and Klausner go on to describe the switching costs effects they find as “[i]nternal learning and
network benefits.” Id. at 752-53.
The concept of network effects within a firm is certainly far removed from
standard network theory. Indeed, as noted above, the problem of network formation for strong network goods such as
fax machines was originally solved by internalization within large firms. We would disagree, however, with the idea
that intra-firm contracting patterns that ran counter to market standards could be considered a network effect of any
kind or, indeed, an externality of any sort.
453.
Id. at 753.
++158
CALIFORNIA LAW REVIEW
of intra-firm learning curves, which is what Kahan and Klausner's data imply, such adherence will
tend to produce heterogeneous terms, not networks.454
6. Implications for Corporate Governance
Even from Klausner, the implications of network theory for corporate law are not sweeping.
Klausner advocates a reconceptualization of the contractarian paradigm because, in his view, the
presence of network externalities undermines the a priori case for presuming that contracts actually
formed in the market are optimal. For the reasons set forth above, we see little reason why the
economic influences Klausner identifies as network effects put much of a dent in the contractarian
454.
Indeed, this is essentially the pattern Klausner hypothesized as suboptimal prior to the adoption of the
ABA model debenture indenture. Similarly, Kahan and Klausner may be read to argue only that marketing effects
cause suboptimal contracting by creating unrealized value, rather than by determining which contract terms a firm
will adopt.
In this sense, such externalities as might exist would be causal in the same sense that both a
pedestrian and a motorist are the cause of any accident between them: even if firms used the a variety of debt terms
such variety might be optimal absent potential, unrealized increasing returns. This usage would be to some degree
problematic because, unlike the motorist and pedestrian, the decision to adopt a term is effected by the term’s value.
The logic of network theory implies that firms would be attracted to uniform terms by the possibility of enhanced
value. If firms do not gravitate toward uniformity, that would imply either imperfect knowledge of the potential for
uniformity to create value or that such value as could be had through network effects was lower than the cost of
switching. Given the safe assumptions that service providers such as bankers and lawyers are generally aware of the
value of uniformity, are able to provide at least some measure of coordination among firms, and have some parochial
interests that would drive them to favor uniformity themselves, the absence of uniformity in the context of an
example such as event-risk bonds is an independent reason to question whether network effects either exist or are in
any way material.
[Vol.
YEAR]
SHORT TITLE
159
presumption. Professor Klausner's principal recommendation for corporate law involves reconceptualizing the contractarian paradigm, gravitating away from the notion of corporate statutes as
unitary products of a hypothetical bargaining process and toward a notion of statutes as the product
of a standard-setting process. Klausner thus suggests drafting corporate statutes to include menus
of possible terms rather than a single majoritarian (or plurality) default term.455
Our analysis in the preceding sections suggests that whatever network effects exist in corporate governance or contract terms are inframarginal, with the exception of interpretive effects,
which are questionable in their own right and of indeterminate implication. We thus see no reason
to suppose that network theory compels the conclusion that corporate contracting is systematically
suboptimal or that menus are necessarily to enhance the value of corporate contracts.456
By the
same token, however, we see no reason actively to oppose the menu concept.457 The only real
downside risk would appear to be that the availability of various items on the menu would induce
too much diversity where uniformity would be preferable. If we are right to suspect that network
455.
See Klausner, supra note 5, at 839-40.
456.
Klausner states that “for terms with significant network externalities, the difference between default
and mandatory terms is more one of degree than one of kind.” Id. at 837 n.242. As we do not read Klausner to
conclude that any of the effects he identifies are in fact significant with respect to any particular term, there seems
relatively little reason to question whether statutory defaults are more mandates than options.
457.
Some existing statutes include menu-like provisions, such as Section 204 of the California
corporation law, Cal. Corp. Code § 204 (West 1990), which provides for optional provisions in articles of
incorporation (as opposed to the provisions required by section 202, Cal. Corp. Code § 202 (West 1990)) and lists
several options. We read Klausner’s proposal, however, to be more substantive and to provide more explicit choices
than are reflected in section 204.
Cf. Robert Rasmussen, Debtor’s Choice: A Menu Approach to Corporate
Bankruptcy, 71 Tex. L. Rev. 51 (1992) (proposing menu options for coprorate bankruptcy).
++160
CALIFORNIA LAW REVIEW
effects exist weakly in this area are weak, this risk seems slight. Indeed, our analysis suggests
that, given the likely inframarginal nature of service network effects, competition among lawyers
and bankers, and the often decisive role they play in choosing terms, there is no clear reason why
law firms and investment banks would not already offer menu alternatives were it efficient to do
so.
Klausner also examines the question of corporate federalism in light of network economics.
He suggests that network effects may explain why Delaware remains the jurisdiction of choice for
large firms when so many other states mimic Delaware's law.458 Couching the question as one of
the costs of remaining incorporated in Delaware or reincorporating elsewhere, 459 the networks argument adds to the debate interpretive effects (the opportunity costs of future clarification of governance terms) and legal service effects.460 For the reasons explained above, we doubt legal service
effects create material efficiencies; to the extent laws of other states mimic Delaware, the marginal
cost of learning such laws will be minimal and, given competition among firms and the inframarginal nature of such effects, likely internalized.461 The desire to avoid such costs is therefore unlikely
to play much of a role in maintaining Delaware's lead.
458.
See Klausner, supra note 5, at 842.
459.
Professor Black argues that reincorporation costs are low, Bernard A. Black, Is Corporate Law
Trivial?: A Political And Economic Analysis, 84 Nw. U. L. Rev. 542, 558 (1990), while professor Romano
believes they are material, Roberta Romano, The Genius of American Corporate Law 34-35 (1993).
460.
See Klausner, supra note 5, at 843-44.
461.
The possibility remains, however, that the parochial interests of the Delaware bar play an important
role in ensuring that Delaware remains responsive to the needs of managers who choose a firm’s state of
incorporation. See Jonathan R. Macey & Geoffrey P. Miller, Toward an Interest-Group Theory of Delaware
Corporate Law, 65 Tex. L. Rev. 469 (1987).
[Vol.
YEAR]
SHORT TITLE
161
The role of opportunity costs from interpretive effects is more difficult to judge. The first
problem is again the relative strength of inherent and network effects. Because the existing case
law is an inherent benefit of a Delaware charter,462 firms have a strong incentive to incorporate in
Delaware on that basis alone. Moreover, as noted above, it is unrealistic to expect that firms will
price the value of future Delaware precedents (network benefits) without significant reference to the
past behavior of Delaware courts (inherent benefits) as an indicator of that value. Klausner is right
to contend that if interpretive network benefits exist, they will extend through most if not all firms
adopting such charters, and thus that the large number of firms presently incorporated in Delaware
may represent a benefit other states cannot replicate merely by mimicking Delaware law.463 But
given the difficulty of distinguishing between decisions based on extant precedent and those based
on interpretive network effects, our intuition that firms themselves do not draw such a distinction,
and the degree to which corporate attorneys influence a firm’s choice of its state of incorporation
and may do so for their own parochial reason, there is no reliable way to conclude that firms would
alter their incorporation choices if interpretive network benefits were absent.
462.
As Klausner states,
[T]he network benefits associated with a contract term consist of future judicial interpretations. Existing case law is
an inherent benefit and can, at least in theory, be incorporated into a contract term explicitly or by reference.
Delaware’s current stock of precedents is thus an inherent benefit of the Delaware charter, not a network benefit. The
network benefit of the Delaware charter is the present value of future judicial interpretations and depends directly on
the number of firms incorporated in Delaware.
Klausner, supra note 5, at 844 n.263.
463.
See id. at 845.
++162
CALIFORNIA LAW REVIEW
G. Contract Terms
As the discussion in the previous section makes clear, the analysis of network theory with respect to corporate law—at least within the contractarian paradigm—overlaps in significant part with
the analysis of contract law. Theories of standardized terms in corporate contracts are largely generalizable to other sorts of contracts; so too are the shortcomings of these theories. However, for
purposes of network theory, we will distinguish between form agreements and negotiated agreements. As Avery Katz has pointed out, form agreements may be conceptualized as an input into
any given exchange that lowers the costs of transacting by avoiding the cost of negotiating over
recurring terms.464 The savings are particularly large among repeat players. Form agreements
might exhibit the sort of interpretive network effects discussed above with respect to corporate
charter terms. The adoption of the Uniform Commercial Code, albeit with material modifications
by various states, reflects a recognition of potential positive interpretive network effects.
As discussed above, however,465 from a lock-in perspective, there is little reason to expect
strong network effects in contract terms. Specific terms will have high inherent value that can gain
little from additional interpretations of the contract. More open-ended terms may gain some clarity
through adjudication, but such clarity will be constrained by the increased number of variables introduced by broader terms and the relatively few number of cases that yield reported decisions.
For example, consider one of the most standardized (indeed, mandatory) “default” rules in contract
464.
Avery Katz, Standard Form Contracts, in The New Palsgrave Dictionary of Law and Economics
(forthcoming 1998); see also Hill, supra note 368, at 2 (“Over time, the form [contract] evolves, mostly for the
better;” but noting the persistence of bad contract language).
465.
See supra notes 414-416 and accompanying text.
[Vol.
YEAR]
SHORT TITLE
163
law—the implied covenant of good faith and fair dealing.466 The implied covenant is an openended term that effectively establishes a totality of the circumstances test, which by definition produces different outcomes in different circumstances. Indeed, despite abundant litigation on this
single contract “term,” few would argue that its meaning is certain today. Even the nature of the
circumstances that are relevant varies from case to case. What is fair conduct for dealing in durable
goods may not be fair conduct for goods with shorter lives, and the standard imposed on vendors
in certain industries (i.e., insurance) may be stricter than the general standard.467 Fact-specificity is
a significant limit on the “certainty” supplied by common contract terms; ask any contract litigator. 468
Furthermore, firms may adopt certain standard terms for a variety of reasons, none of which
necessarily imply the existence of network effects. Financial contracts, for example, are negotiated
as a whole, and parties seeking agreement on an important point may not want to expend negotiating capital to revise relatively unimportant terms. 469 If a standard term imposes an opportunity cost
on one party, that cost may be recouped through other terms more favorable to the party. It would
466.
See, e.g., UCC § 1-203.
467.
Similarly, from a governance perspective, conduct that might satisfy a fiduciary duty to corporate
shareholders (or creditors, in the case of insolvency) might not satisfy such a duty in the case of a limited
partnership, with its different taxation structure and different doctrinal traditions.
468
Kahan and Klausner do not suggest otherwise. See Kahan & Klausner, Path Dependence, supra note
353.
469
See Hill, supra note 368.
++164
CALIFORNIA LAW REVIEW
then be difficult to conclude that the contract as a whole was suboptimal.470 A similar cost/benefit
analysis will apply to most negotiated contract terms.
Both within corporate and contract law, courts must confront the question of how to facilitate
formation of efficient contracts at the lowest cost. As noted above, there is little reason to expect
adherence to a single suboptimal form, at least for relatively large transactions in which firms are
advised by sophisticated intermediaries. If we consider securities as simply one form of contracts,
the market for contract terms is a highly dynamic one in which new products such as derivatives
are commonly introduced to compete both with one another and with more traditional investments.471 Again, it is not clear that network effects have much to offer in this market. In the presence of weak network effects one would expect intermediaries such as investment bankers to have
a variety of alternatives to offer, each of which enjoyed some degree of network effects. In the
presence of stronger effects, one would expect menus of alternative terms to be superfluous. As
with corporate law generally, we would expect at most modest network effects subject to significant constraints, so that alternative menu-based systems would have little discernible effect on the
contracts actually drafted in the market.
470.
We can attest from anecdotal evidence that parties will sometimes insist on the inclusion of language
accepted by reputable investment banking institutions in previous deals because such language is marginally
beneficial to a party and its prior acceptance by a reputable intermediary provides a credible negotiating position.
The other party to the transaction may have other deal points it considers more important, and therefore may choose
not to fight over the “standard” provision in order to obtain its preferred language in some other portion of the
agreement. This type of logrolling over individual contract terms may potentially perpetuate standard language in
certain instances, but does not necessarily reflect suboptimal contracting.
471
See Henry T.C. Hu, Hedging Expectations: “Derivative Reality” and the Law and Finance of the
Corporate Objective, 73 Tex. L. Rev. 985, 988 (1995).
[Vol.
YEAR]
SHORT TITLE
165
Contracts of adhesion—those drafted entirely by one party and offered on a take-it-or-leave-it
basis—deserve separate treatment. If one party cannot bargain over contract terms, adhesion contracts will likely standardize around terms favorable to the drafting party.472 Indeed, since the only
means of “bargaining” available to non-drafting parties is the threat to go to a competitor offering
better terms, drafting parties may have a strong incentive to standardize their terms to squelch competition. Friedrich Kessler has referred to such uniform, cartel-like terms within an industry as
“private legislation.”473 This might be thought of as a “network effect” of a sort—those who draft
contracts of adhesion with one-sided terms benefit if their competitors adopt the same one-sided
terms.474 Unless the term itself is so important to the other side that they are willing to forgo any
transaction in the market, they will be stuck with the uniform term.
472.
Rubin cites as an example self-help terms imposed by vendors and lenders. The existence of strong
cost barriers to litigating such terms helps perpetuate them, even in circumstances where they would probably be
unenforceable if taken to court. Rubin, supra note 395, at 125-31.
473.
Friedrich Kessler, Contracts of Adhesion—Some Thoughts About Freedom of Contract, 43 Colum.
L. Rev. 629, 640 (1943) [Professor Lemley, doesn’t seem that Kessler used exact term “private
legislation.” Still OK?]. See also Mark A. Lemley, Shrinkwraps in Cyberspace, 35 Jurimetrics J. 311, 31920 (1995); Robert P. Merges, Intellectual Property and the Costs of Commercial Exchange—A Review Essay, 93
Mich. L. Rev. 1570, 1611-12 (1995).
474.
In fact, however, this effect most resembles a traditional cartel in which all participants have
individual incentives to cheat and a collective incentive not to cheat. The benefits to cartel participation do in fact
depend on how many other existing members of the industry join the cartel. On the other hand, a cartel is unlike a
network in that expanding the number of participants in the market is a bad thing rather than a good thing for
existing firms.
++166
CALIFORNIA LAW REVIEW
In this model, standardization becomes the desirable outcome for the drafting parties, but not
necessarily the optimal outcome for society. Where private standardization eliminates competition
over contract terms, mandatory (or at least stricter) statutory default rules may serve as a viable alternative. Statutory defaults would lead to more socially optimal contracts if (but only if) they include contract terms with higher social values than those selected through private standardization.
Although we have not undertaken a comprehensive study, we suspect there will be at least a few
cases in which this would be the likely outcome. Unfortunately, the modern legislative trend appears to be in the opposite direction. Draft UCC Article 2B,475 dealing broadly with transactions in
“information,” including intellectual property, would adopt nearly no mandatory rules, and would
make it much easier to change default rules through contracts of adhesion (or even without the express agreement of the other party).476
Contract law also figures into debates over network theory in another way—as a potential
remedy to problems caused by network effects. Contracts underlie virtually all of modern commerce, including commerce in network markets. As the Supreme Court's Kodak opinion points
out, when antitrust aficionados express confidence that "the market will fix” some anticompetitive
problem, they are implicitly invoking the operation of contract law and thus the problems inherent
in the process of contracting.477 Thus, the mechanism of contract implements many economic
475.
The May 1997 draft (the most recent one) is available from the University of Houston Web server
(visited Aug. 4, 1997) ‹http://www.law.uh.edu/ucc2b/050597/0505_2b.html›.
476.
See UCC 2B §§ 112, 308(e) (May 1997 draft). See also David A. Rice, Digital Information as
Property and Product: U.C.C. Article 2B, 22 U. Dayton L. Rev. 621, 643-45 (1997) (criticizing this aspect of the
draft).
477
Eastman Kodak Co. v. Image Technical Services Inc., 504 U.S. 451, 479 (1992).
[Vol.
YEAR]
SHORT TITLE
167
strategies relating to networks. For example, firms that seek to predate in order to acquire market
share, often do so through agreements. Additionally, firms may seek to draft intellectual property
licenses that indirectly contribute to their market share—witness Microsoft's former practice of
charging a license for its operating system to hardware manufacturers based on the number of
hardware units sold, regardless of whether Microsoft's software was installed.478
One might attempt to resolve these problems within contract doctrine itself, rather than intervening with antitrust law or changing intellectual property rules. Various contract doctrines—and
certainly contract’s normative principles— might prove the best way to address some of the problems present in network markets. For example, as noted above, network markets are particularly
susceptible to consumer expectations. Because consumers prefer to own the standard product, an
announcement by the standard-setting firm might cause consumers to forgo purchases of other
products in anticipation of the new standard. If the announcement turns out to fall into the category
of "vaporware," an announcement of a product might induce reliance that, with proper analysis and
limitation, could lead to a cognizable claim under contract-based theories.479 Firms that announce
open architecture in order to induce the production of complementary goods that help the firm establish itself as a standard could be estopped from “closing” future iterations of their software in
the event third parties invested capital and brought value to the network as a whole through the
creation of complementary goods.480
478.
See supra notes 72-78 and accompanying text (discussing challenges to this practice).
479.
Cf. supra notes 92-96 and accompanying text (treating such claims under antitrust law).
480.
One of the authors has previously argued for such a result in the context of the Data General case. See
McGowan, supra note 10, at 836-841.
++168
CALIFORNIA LAW REVIEW
We hasten to say that these are merely illustrations. Each requires much further analysis and,
with the exception of the examples we discuss below, that analysis is beyond the scope of this article. The point remains, however, that one cannot speak intelligently about remedies to problems of
network theories without considering contract law. We will return to this point in Part III.
III
Deriving Principles for Judicial Treatment of Network Effects
A unifying them in these cases is disunity. As we observed in Part II, “network effects” do
not constitute a single phenomenon. Rather, different types of network effects exist, and they vary
in strength. Courts and legislatures faced with claims of network effects must have some mechanism for distinguishing “good” from “bad” arguments (in the positive sense of identifying whether
a network effect really exists), and for distinguishing “good” from “bad” effects (in the normative
sense of deciding whether a particular network effect is beneficial or harmful, and what if anything
should be done about it). To complicate matters even further, the work of Michael Klausner and
others suggests that the legal decisionmaking process itself may contribute to the formation of networks.481
We begin with what we believe is generally accepted as an example of successful adaptation of
economic theory to the law: the “fraud on the market” doctrine in securities law. The Supreme
Court in Basic, Inc. v. Levinson adopted the efficient market hypothesis, creating a rebuttable presumption of reliance for purposes of ruling on class certification in securities cases involving
stocks trading in efficient markets. 482 Commentators widely endorse this presumption of reliance
481.
See supra notes 353-463 and accompanying text (discussing Klausner’s argument).
482.
Basic, Inc. v. Levinson, 485 U.S. 224, 247 (1988). As Professor Fischel has put it, the fraud on the
market doctrine “has been of particular interest to scholars because courts have expressly based their acceptance of the
[Vol.
YEAR]
SHORT TITLE
169
with respect to the class certification issue.483 Professors Macey and Miller rightly point out, however, that courts ought instead to ask whether the market for a stock functions efficiently with respect to the specific information involved in the case before it, rather than whether the market for
the stock functions efficiently in general.484
What accounts for this apparently successful instance of adaptation? We suspect the answer
lies in the relative familiarity of judges with the class certification problem. Federal Rule of Civil
Procedure 23(b)(3), the provision most commonly invoked in these cases, requires that common
questions of law or fact predominate among the claims of the individual class members.485 A plaintiff alleging fraud through misstatements must demonstrate reliance on the relevant statement.486
Adducing such proof from each individual plaintiff would effectively destroy the cost advantages
of class actions, and perhaps the ability of class actions to fulfill their essential function of surmounting collective action problems.
theory not on the legislative history of the securities laws but rather on the academic support for the efficient
markets hypothesis.” Daniel R. Fischel, Efficient Capital Markets, The Crash, And The Fraud On The Markets
Theory, 74 Cornell L. Rev. 907, 907 (1989) (citing, e.g., In re LTV Sec. Litig., 88 F.R.D. 134, 144 (N.D. Tex.
1980). This is not to say that courts did not certify classes in securities litigation before the efficient market
hypothesis was developed. It is fair to say, however, that neither the Exchange Act nor the Federal Rules of Civil
Procedure provide the basis for that presumption offered by the efficient market hypothesis.
483
See Basic, 485 U.S. at 991 (“Commentators generally have applauded the adoption of one variation or
another of the fraud-on-the-market theory.”).
484.
See Jonathan R. Macey & Geoffrey P. Miller, Good Finance, Bad Economics: An Analysis of the
Fraud-On-The-Market Theory, 42 Stan. L. Rev. 1059, 1083 (1990).
485
See Fed. R. Civ. Pro. 23(b)(3).
++170
CALIFORNIA LAW REVIEW
Enter the “fraud on the market” theory, which implied that plaintiffs could have suffered the ill
effects of statements they never saw or heard. The theory appeared in the cases very quickly following its ascendance in the field of financial economics.487 By 1975, the Ninth Circuit in Blackie
v. Barrack felt able to assert that "[a] purchaser on the stock exchanges . . . relies generally on the
supposition that the market price is validly set and that no unsuspected manipulation has artificially
inflated the price . . . whether he is aware of it or not, the price he pays reflects material misrepresentations."488 The decisions emphasized the practical economics of securities litigation: if no class
were certified, the court would probably not reach the question of whether any fraud had been
committed. From the courts' perspective, if no fraud had been committed, they could end the case
on summary judgment. And if a triable issue existed, no court was going to bring thousands of
individual investors in to testify about their investment decisions.489
486.
See Basic, 485 U.S. at 243.
487.
The initial case was Blackie v. Barrack, 524 F.2d 891 (9th Cir. 1975).
The efficient market
hypothesis or “random walk” theory itself may be traced back at least to 1965, with Eugene Fama’s Random Walks
In Stock Market Prices, 21:5 Fin. Analysis J. 55, 55-59 (1965). Fama’s influential 1970 article, Efficient Capital
Markets: A Review of Theory and Empirical Work, 24 J. Fin. 383 (1970), reflected growing interest in and
agreement with the theory within its field. See Jeffrey N. Gordon & Lewis A. Kornhauser, Efficient Markets,
Costly Information, And Securities Research, 60 N.Y.U. L. Rev. 761, 762 n.1 (1985) (noting influence of Fama’s
1970 article). For a theory to achieve prominence in its own field in five years is notable; for it to be adopted by the
law five years later is quite remarkable.
488.
Blackie, 524 F.2d at 907. See also Richard W. Jennings, Harold Marsh, Jr., & John C. Coffee, Jr.,
Securities Regulation 1241-45 (7th Ed. 1992) (tracing history of adoption of fraud-on-the-market presumption).
489.
We wish to be clear that what we set forth in the text is intended as our view of likely judicial
attitudes on the class certification question. Whether such attitudes increase social welfare insofar as it relates to
[Vol.
YEAR]
SHORT TITLE
171
Thus, while the theory was relatively new even in its field of origin, it fit well with the intuitive desire of judges to minimize costs, avoid duplicative suits, and resolve cases on the merits.
Judges charged with running a docket were quick to grasp the sensible, practical implications of
the theory, and thus welcomed the resulting rule. It is firmly established in the law today, even
though commentators have since called into question the economic theory on which it relies.490
We may safely presume judicial expertise in matters of judicial administration, and reliance on
that presumption may quell most fears that appealing but vacuous new theories of court procedure
will mislead judges. For questions of substantive law, however, the process of legal adaptation
involves greater risk. Here, the fraud on the market theory offers a cautionary note as well. Its
adoption has essentially mooted many of the elements of the substantive cause of action under Rule
10b-5, reducing them at best to a battle between experts. As Professor Fischel puts it,
[T]he fraud on the market theory has profound implications for securities litigation
far beyond reformulation of the reliance requirement. The fundamental question in
securities fraud litigation is whether a challenged disclosure has caused an investor
to suffer an injury and, if so, by how much. . . . Indeed, inquiry into materiality,
securities litigation is beyond the scope of this article; suffice it to say that the view described in the text is not selfevidently efficient.
490
See, e.g., Gilson & Black, supra note 401, at 117-33 (discussing debate over the continuing vitality
of the capital asset pricing model, on which efficient market theory rests); Donald C. Langevoort, Theories,
Assumptions, And Securities Regulation: Market Efficiency Revisited, 140 U. Pa. L. Rev. 851, 854 (1992) (noting
criticisms of the assumptions on which the theory rests, with particular emphasis on noise trading theory, and
arguing that a “gulf . . . has developed between the current economics literature and the persistent, seemingly static,
conception of market efficiency in the legal culture”).
++172
CALIFORNIA LAW REVIEW
causation, and damages, as well as reliance, is subsumed into the question whether,
and to what extent, a fraud on the market has occurred.491
While Fischel is right, the Supreme Court in Basic, Inc. v. Levinson held that the fraud on the
market doctrine could be invoked as a rebuttable presumption for class certification and that the old
test for materiality developed under the proxy rules492 applied to 10b-5 cases, without considering
the tension between the two rulings.493 Courts and Congress have spent countless hours in the
years following Basic attempting to prevent a reasonable procedural doctrine from altering the
substantive law.494
Nobody said adaptation would be easy. Courts and legislatures faced with arguments of this
type need heuristics to guide them in decisionmaking. In this Part, we suggest factors courts
might use to evaluate a claim that a market (or the law itself) is characterized by network effects. It
is of course impossible to offer anything like a full taxonomy of network effects claims. The recent nature of network economic theories and the continuing debates among economists about their
implications show that the economics of networks is itself still under construction. Further, we do
not purport to have identified all or even most of the potential roles network effects may play in
491.
Fischel, supra note 482, at 909. See also In re Clearly Canadian Sec. Litig., 875 F.Supp. 1410,
1414-15 (N.D. Cal. 1995) (holding that upon proof of misstatement or ommission, coupled with presumption of
reliance, the elements of materiality, causation and damages were also satisfied).
492.
TSC Industries, Inc. v. Northway, Inc., 426 U.S. 438, 449 (1976) (information “is material if there is
a substantial likelihood that a reasonable shareholder would consider it important in deciding how to vote”).
493
See Basic, 485 U.S. at 249-50.
494.
The most recent result is the Private Securities Litigation Reform Act of 1995, Pub. Law No. 104-
67, 109 Stat. 737 (1995) (codified as amended at 15 U.S.C. §§ 77-78).
[Vol.
YEAR]
SHORT TITLE
173
legal argument. Nonetheless, in the pages that follow we offer a number of yardsticks courts
might use to measure such claims.
A. Nature and Strength of the Effects
A court presented with an argument based in network theory must scrutinize the claimed effect
closely. As we have seen, a wide variety of different network economic effects can be lumped under the label “network externality.” Even before analyzing evidence of market factors, a court confronted with a network effect argument must determine the type of network effect at issue. An
“actual” network, such as the interconnection effects associated with telephone networks, is much
stronger than a “virtual” network or than a general positive feedback relationship. This difference
in strength manifests itself in various ways. First, actual networks are much more likely than virtual networks or positive feedback markets to converge on a single standard, and that standard is
more likely to be universal. Network effects are not always absolute; sometimes multiple products
can each build a core of users with its own partial network effect. For example, despite the dominance of the Microsoft operating systems, there remains a core group of dedicated Macintosh users
who are locked in to their own (losing) network standard. Similarly, Betamax owners kept Betacompatible videos in stores long after it was clear that Beta would not be the dominant VCR standard. But this sort of coexistence is less and less likely as the strength of the network effect
grows. Economic factors may incline consumers to choose the Web browser that most others are
choosing, but they can also resist that inclination if they have a good enough reason. They are
much less likely to be able (or willing) to buck the trend toward a universal telephone network by
joining an incompatible rival network.
Second, a court must also focus on the strength of the network effect. The benefits associated
with being in the winning network generally increase with the strength of the network effect. Existing users of the telephone network receive more benefit from the addition of a million people to
++174
CALIFORNIA LAW REVIEW
that network than existing users of Web browser software would receive from the proportional addition of more users of the software. For both of these reasons, strong networks have a stronger
lock-in effect, are more durable, and take greater force to displace than weak ones.
With respect to actual or virtual networks, these facts alone do not unambiguously imply any
particular legal conclusion. A strong network effect can be either good or bad, depending on the
nature of the standard, the difficulty of transitioning to alternative standards and, from a legal perspective, the effect the standard has on the goals of the principal fields of law with which it intersects.
Strong networks can even be both good and bad, as where a destructive standards
competition is resolved in favor of a suboptimal product that nonetheless gives consumers the
benefits of the network effect.
Nonetheless, we can draw some tentative conclusions about the strength of network effects.
Because strong network effects tend to produce more uniform standards with more durable effects,
the law should be more concerned with attempts to control or exclude others from participating in a
standard if the standard is part of an actual or strong virtual network than if the battle is merely over
convenience effects. At the same time courts should be less willing to interfere with the operation
of the strong standard itself, since adoption of strong network standards generally has large positive network effects.
We believe that courts should substantially discount network claims based only on evidence of
positive returns to a certain level of scale. For example, we are unaware of any compelling evidence that suggests that the law should be revised to take care of the problems faced by the only
owner of an exotic car in a small town.495
495.
If I am the only person who owns a Range Rover in a small training will not be covered by the
minimal business I am likely to generate. On the other hand, if 100 people in the town own Range Rovers,
[Vol.
YEAR]
SHORT TITLE
175
Courts should be particularly cautious in cases where the network does not involve the standard consumer goods on which most of the economic literature has been based. One may draw
analogies between telephone networks and contract terms or corporate statutes. Such analogies are
not complete, however, and as Klausner has noted, they require further study before confident
conclusions can be drawn.496 If an analytical approach to a finding of intra-firm network effects,
for example, it is likely that some other economic force is actually at work. To take another example, many markets subject to network effects involve the combination of distinct vertically-related
products, such as VCRs or software, to satisfy consumer needs. Corporate governance terms
themselves, however, are the products of vertical relationships between firms and intermediaries,
with potential network effects operating at both levels. These and the numerous other variables
discussed above suggest that courts and legislatures should tread especially lightly in adapting network theory to fields in which the markets themselves are not well understood.
B. Evidence Supporting Network Effects
Evaluating a claim of network effects involves not only an inquiry into the type of network
effect at issue, but also a factual determination of whether such a network effect actually exists. In
specializing in Range Rovers may make economic sense. One might think of this as a “network effect—the other
99 people who bought Range Rovers collectively conferred a benefit on existing Range Rover owners (i.e. me). But
this is not a network effect, precisely because it is tied directly to the economies of providing service, and does not
extend over the range of market demand. The marginal customer who warrants the addition of a garage has benefited
all the existing customers, but other marginal added Range Rover customers do not necessarily confer any such
benefit. Indeed, customers beyond 100 may have a negative effect on existing owners, causing the garage to be too
busy.
496.
Klausner, supra note 5, at 775.
++176
CALIFORNIA LAW REVIEW
part, this determination does not differ from the evaluation of any legal claim: the factfinder must
review the evidence to determine whether the assertion that a network effect exists is real or illusory. This evidence will undoubtedly be economic in nature, and proffered evidence may in fact
consist of economic expert testimony, or even anecdotes. Courts have dealt with such economic
evidence in other contexts, and the same general principles apply here. However, economic evidence contains traps for the unwary. The recent challenges to the once widely-accepted QWERTY
story497 offer some reason for caution: economic explanations for historical events do not come
free from doubt, and there is always a danger that particular circumstances will be misinterpreted as
involving network effects. In addition, there are some more general pitfalls associated with factual
claims that a market is characterized by network effects. Many economic phenomena may resemble network effects at first glance. In this section, we identify four such phenomena that should
not be mistaken for network effects of any kind, not even the positive feedback effects discussed
above.
First, general positive externalities often resemble network effects. In economic terms, a
positive externality exists whenever a private action confers uncompensated benefits on third parties. Education exhibits certain positive externalities; while the student obtains some benefit from
497.
One of the paradigmatic examples of the pernicious effects of networks was the QWERTY typewriter
keyboard. According to the classic story, the inferior QWERTY keyboard layout has dominated the market for over a
century because of network-related lock-in: no one wants to be the first to switch to an alternate keyboard. See,
e.g., Paul A. David, Clio and the Economics of QWERTY, 75 Am. Econ. Rev. 332 (1985). In several articles,
Liebowitz & Margolis have cast doubt on the historical basis of the claim that the alternate Dvorak keyboard is in
fact better. See S.J. Liebowitz & Stephen E. Margolis, The Fable of the Keys, 33 J.L. & Econ. 1 (1990);
Liebowitz & Margolis, Uncommon Tragedy, supra note 5, at 133. For efforts to rehabilitate the Dvorak story, see
Jared Diamond, The Curse of QWERTY, Discover, Apr. 1997, at 34.
[Vol.
YEAR]
SHORT TITLE
177
her education, the full social benefit of her education is more diffuse, including the effects of her
productivity, information she passes on to others, the fact that she is less likely to commit crimes,
and so on. Because she cannot capture or charge for these benefits, the student’s private return to
education is less than the social return, and she will not “consume” enough education unless some
steps are taken to compensate for this externality. Education is not a network effect, however.
Whatever social benefits my education confers are not a function of the number of other people
being educated.
Some general positive externalities may at first glance appear to be network effects. For example, building an attractive destination store in a shopping mall confers benefits on other stores in
that mall because it brings more customers in to shop. This is a positive externality, but not a network effect. While shop owners would all benefit from increased quality stores in the mall up to a
certain point, the benefit does not extend across the entire range of market demand. Stores in competition with the new ones may lose rather than gain. To the extent the mall is fixed in size, new
stores may displace rather than coexist with existing stores. Even assuming no external constraint
on mall size, the mall may have a maximum efficient size—a mall which is too big for consumers
to handle easily will not confer benefits on its members.
This leads us to a second, related economic phenomenon sometimes confused for network
effects: economies of scale. Economies of scale exist when the cost of production per unit declines as the number of units produced increases.498 Economies of scale frequently occur in industries characterized by high fixed costs. 499 If it costs $100 million to build a new steel factory, its
498.
See Pindyck & Rubinfeld, supra note 10, at __ [recommend pincite].
499.
See generally Blair & Kaserman, supra note 89, at 94-95. But cf. Herbert Hovenkamp, supra note 52,
§ 6.3, at 247 (1994) (courts can’t effectively determine when scale economics exist).
++178
CALIFORNIA LAW REVIEW
owner will want to use that factory for more than producing a single steel bar. The marginal cost
of producing each new bar may decline up to a certain point, as fixed costs are amortized. Economies of scale resemble network effects because both phenomena exhibit increasing returns over
significant portions of the demand curve. The key difference lies in the source of the returns. In
actual networks markets, each additional consumer adds value to the network by increasing the
number of people who can be reached through the network. For industries displaying economies
of scale, by contrast, the effect is on the supply rather than the demand side. The increased value
comes from the increased productivity of capital investments, regardless of the manner in which
consumers use the good in question.
These scale economies are distinct from network effects. Microsoft enjoys very large economies of scale in producing its software because its fixed costs are quite high and its variable costs
quite low.500 If Microsoft's products could be used only by individual consumers for individual
purposes—if they did not interoperate either with like goods or with compatible goods—Microsoft
would still enjoy significant economies of scale simply by virtue of this cost structure. The interoperable nature of its products adds value in part made possible by the same consumer demand
that facilitates scale economies—both the ability to share files and the size of an aftermarket of application programs will increase with demand—but the value created by the network benefits of
interoperability is distinct from the value created by declining marginal costs of production. To
distinguish a genuine network phenomenon from traditional economies of scale, a court should
500.
The “marginal cost” of a new copy of Microsoft Windows is the cost of the disk (a few cents), the
labor required to copy it (essentially none), and the cost of the manuals, packaging, and distribution of the box itself
to consumers (by far the largest portion). Software distributed on the Internet has essentially no marginal cost, as it
can be downloaded by the consumer on their own time, albeit with some commitment of computer resources by the
supplier.
[Vol.
YEAR]
SHORT TITLE
179
seek to ascertain the portion of a firm's returns generated by the cost structure of the industry in
which it operates—economies that end at the point the product is shipped—and the portion generated by the manner in which consumers use or seek to use the product—returns generated by the
consumers themselves or perhaps by other firms making complementary goods, and which begin
only after the product is shipped.
Classical economic theory presumes that economies of scale at some point become diseconomies of scale, producing the classic U-shaped cost curve.501 This assumption will not always hold
true, however. Classical theory treats markets in which returns increase (marginal costs fall) over
the entire range of market demand as instances of natural monopoly, in which efficiency dictates
that a single firm serve the entire industry. In such cases, additional consumers in the industry
benefit existing consumers by amortizing a greater share of the fixed cost. 502
Some of these
“natural monopolies” do indeed show up in the areas we have been discussing, such as telephony. 503 But the scale economies in telephony are distinguishable from the network effects. It may
be efficient for a single firm to build the local telephone distribution network for a city, but that
firm will not necessarily be the most efficient supplier of telephone wiring worldwide.504 Further,
once the wire is installed, many different companies might compete in providing telephone
switching services in the city. And different cities may efficiently hire different firms to provide
their telephone wiring and switching. If economies of scale were all there was to the telephone in-
501.
See Samuelson & Nordhaus, supra note 292, at __ [recommend pincite].
502.
On this phenomenon, see Lemley, supra note 35, at 1054-55.
503.
See supra notes 286-305 and accompanying text.
504
Even this assumption has been challenged of late, as new technologies such as cable and wireless
service seek to compete with the existing twisted-pair telephone wire. [cites -- need to ask Lemley]
++180
CALIFORNIA LAW REVIEW
dustry, we would expect to see numerous competitors (and no government regulation). But because of network effects, each of these competitors must link to the same global telephone network
in order for consumers to maximize the value they get from the network. Different companies can
build the network—economies of scale don’t prevent that—but they can’t build different networks.
The distinction between economies of scale and network effects is important to understand
because network effects are significantly more limited than economies of scale. Those who are
unclear on the distinction occasionally suggest that network effects are or will be ubiquitous, even
in standard products markets.505
Third, certain aspects of “path dependence” resemble network effects. Path dependence refers
to the inherent tendency of history to influence present decision-making. Mark Roe cites as a paradigmatic example the construction of a road that is not straight for reasons that were once important
but are today irrelevant. As the road develops over time, the history of the road influences (albeit
unconsciously) decisions about how and where to construct new buildings.506 Path dependence
505
The most notable example is Kevin Kelly, New Rules for the New Economy, Wired, Sept. 1997, at
140, 143 (“In the future, cotton shirts, bottles of vitamins, chain saws, and the rest of the industrial objects in the
world will also obey the law of plenitude as the cost of producing an additional copy of them falls steeply, while the
value of the network that invents, manufactures, and distributes them increases.”). Kelly is wrong. In the first
place, we doubt that a “network” of manufacturers in this sense is a meaningful economic concept. Even if it were,
chain saws simply do not exhibit demand-side economies of scale. We would be quite suprised to see chain saws
given away in the near future by companies hoping to somehow reap an advantage from having more free chain saws
loose in the world. Cf. id. at 188 (“the extension of this logic says that the most valuable things of all should be
those that are given away”).
506.
(1996).
See Mark J. Roe, Chaos and Evolution in Law and Economics, 109 Harv. L. Rev. 641, 643-44
[Vol.
YEAR]
SHORT TITLE
181
may take the form of rational economic decisionmaking—it costs less now to repave the existing
road than to move it altogether—or of a failure to understand sunk costs. In the latter case, society
may not take a rational action because it feels bound by the past.507
The concepts of network effects and path dependence obviously overlap. Path dependence
resembles network theory because both tell stories about “lock-in” to existing standards. Networks will virtually always take shape over time; indeed the term "installed base" is inherently
temporal. The types of externalities potentially created by networks—such as tipping or lock-in
externalities—also will generally appear only over time. Nonetheless, courts should distinguish
between purely historical lock-in effects that exist without regard to adoption decisions, other than
the instant consumer’s earlier choices, and lock-in effects which stem from the demand function of
others. One way to do this is to ask what would happen if the market were suddenly placed at time
1, so that historical factors were rendered irrelevant. The road drive by path dependence might not
end up in the same place, but telephone networks would still tend to converge on a single standard. 508
Finally, one might wrongly assume that exclusionary groups necessarily exhibit network effects. Certainly various non-network groups derive value from membership, which increases with
the number of members—political and social organizations, for example.509 However, not all such
507.
See id. at 651-52.
508
Path dependence—rather than network effects—may explain the persistence of inartful language in
standard form contracts that Hill observes. Hill, supra note 368, at 3.
509.
Cartels are another. Cartel members benefit from the participation of their competitors in the cartel,
since the greater the percentage of competitors who participate, the less vulnerable the cartel will be to cheating.
However, this is not a pure network effect, since cartel members do not actually benefit from the addition of new
competitors to the market, and indeed would prefer to limit the total number of competitors.
++182
CALIFORNIA LAW REVIEW
groups truly exhibit network effects. Economic literature suggests that a relatively small group of
committed people may be more effective in politics than a larger group of less committed people.510
Further, at least some such groups obtain value precisely from limiting membership, either to preserve their presumed exclusivity or to distinguish themselves from “everyone else.” A political
party with 100 million members is more powerful than one with 100 members, but a political
“party” whose members include everyone in the country has no power at all in an important sense.
Groups that derive value from excluding rather than including others are not good examples of
network effects. Indeed, they are in some sense the opposite of a true network.
Of course, for some purposes a court may not care whether a claimed economic effect really is
a network effect or not. If the court seeks to internalize a positive externality, it might choose to do
so regardless whether the externality is a network effect. But we think it important to distinguish
these classes of economic phenomena. In some instances, the law treats network economics in a
strikingly different manner from related effects, like economies of scale. Courts should at a minimum strive accurately to identify the issues confronting them before taking (or abstaining from)
legal action on the basis of a claimed economic effect.
C. Ameliorative Effects
Courts evaluating claims based on network effects should also be cognizant of three factors
that serve to temper the strength of network effects: nonexclusivity, open standards, and the presence of intermediaries to facilitate transition between standards. Certain network markets are less
vulnerable to lock-in than others. In particular, nonexclusive standards may allow consumers to
gain the benefits of widespread adoption without foregoing the benefits of either competition or
510.
See, e.g., Mancur Olson, The Logic of Collective Action (1971); McGowan & Lemley, supra note
163, at 324-25.
[Vol.
YEAR]
SHORT TITLE
183
improved technical standards. For example, ATM networks are largely nonexclusive. There is no
technical reason why a bank can’t link its ATMs to every ATM network,511 giving the bank’s consumers the maximum network benefit and (not incidentally) making it possible for network providers to compete with each other over the long term.512 Nonexclusivity partially counteracts the
potential lock-in effects of even a strong network, since consumers do not have to give up one
standard to embrace another. They do have to embrace the leading standard, however, meaning
that the owner of that standard has potential market power.
For similar reasons, the open nature of a network standard may ameliorate the strength of that
standard.513 Even if network effects force all consumers to migrate to a single product standard,
they (and society) will benefit if numerous companies compete to provide products compatible with
that standard. Not only will the price of the product standard fall, and the adoptions of the stan-
511.
Indeed, most banks are linked to multiple ATM networks, although not to all of them (for reasons
discussed below).
512.
Where consumers can belong to multiple overlapping networks, the switching costs of moving to a
new network approach zero. This has significant implications for the behavior of the networks themselves.
If
networks are to attract members, they must compete on some ground other than simply the number of members. A
network that had the most members, but which offered unattractive terms compared to other networks, could easily
find itself replaced as consumers signed up en masse with its more attractive competitor, and then found that they no
longer needed the first network.
513.
Open standards can exist for a variety of reasons: because the market drives “owners” of standards to
open them up, as was the case with the IBM PC and the VHS VCR; because it is not technically possible to prevent
others from adopting the standard (think of the two-prong electric plug or the design of cars optimized for driving on
the right side of the road); or because the law explicitly allows access to a standard.
++184
CALIFORNIA LAW REVIEW
dard correspondingly rise toward the optimal level,514 but competition within a standard should
spur technological innovation toward improved standards, and particularly toward improvements
that can be reached by migration, rather than by throwing out the old standard entirely. Similarly,
intermediaries may facilitate transition among standards, as in the case of investment banks and
event-risk debt. The same dynamic may diminish the significance of network effects in other markets as well. Some early software programs, for example, were explicitly designed to serve as
bridges between incompatible application programs, translating files between the two.
These ameliorative effects have two important implications for courts considering a network
effects claim. First, both nonexclusivity and open standards reduce the lock-in effects otherwise
exhibited by strong networks. Further, open standards generally reduce the economic distortions
associated with monopoly ownership of standards. The presence of one or more of these factors
therefore lessens the need for judicial action to “correct” or alter the effect of the network market.
In contrast, markets with exclusive or closed standards, or markets in which intermediaries cannot
facilitate transitions for some reason, are more likely to require judicial intervention. In effect,
these tempering factors combine with the nature of the claimed network effect to produce a more
nuanced view of the “strength” of the network effect.515
514.
In a network market, each new consumer benefits all existing consumers. This benefit takes the form
of an uncompensated positive externality—there is generally no effective way for existing consumers to pay back the
additional value they receive. Thus, at the marginal cost to the supplier, not enough new consumers will enter the
market. The socially optimal price for new consumers should be below the marginal cost of serving them, reflecting
the benefits their entry confers upon others.
515
See McGowan, supra note 10, at 847-48 (suggesting that network effects need not be cause for
antitrust concern in circumstances where competitors have legal access to the standard).
[Vol.
YEAR]
SHORT TITLE
185
Second, courts should evaluate claims for judicial relief in part based on the effect they have
on the openness of network standards. Particularly with respect to strong network effects, courts
should favor legal solutions that open standards to competition, either through nonexclusivity or
through group participation in a single standard. Such solutions may not always be appropriate,
nor should they always be judicially compelled. Courts should remain cognizant of the legal environment in which they operate: for example, a court’s declaration that network standards can never
be “owned” as intellectual property rights would likely have significant negative effects on incentives to invent, and perhaps on intellectual property doctrine in general.516 Nonetheless, courts and
legislatures should take account of these ameliorative effects in appropriate cases when setting legal
policy within this broader framework.
D. Significance of the Network Effect in the Industry
Even once a court has established that a network effect operates in a market, determined its
nature and strength, and evaluated any ameliorative effects, it should also asses the significance of
the network effect in the context of the market in which it occurs. Consumers make purchasing
decisions within markets that have product-space, geographic-space, and temporal boundaries.
Networks exert their strongest effects when they coincide with or are broader than consumer markets. This is the case with telephones, for example. Consumers buy both local and long-distance
telephone service. Regardless of where they live, however, or what kind of telephone service they
seek, they want their telephones connected to the single worldwide telephone network. The network effect in this market encompasses the consumer market (and indeed a number of different
geographic markets linked together by the network itself).
516.
See id. at 776.
++186
CALIFORNIA LAW REVIEW
By contrast, not all network effects fill the product-space of a consumer’s market choice.
Consider automobile fuel doors. Automotive designers can’t seem to agree where they belong. In
different model cars, the fuel door might be on the driver’s side, the passenger’s side, or the back
of the car. Drivers of borrowed or rental cars must either invest time in learning the location of
their fuel door, or end up parking on the wrong side of a gas station island. One can make a plausible argument that this is a convenience-related network effect operates and that the market exhibits suboptimal heterogeneity (foregone efficiencies) of gas-door location. Even if this constitutes a
network effect, though, it does not drive the market to standardization, for the simple reason that it
is extremely unlikely ever to influence a purchasing decision in a real market. Consumers don’t
buy fuel doors; they buy cars, and the fuel door contributes only trivially if at all to that purchasing
decision. A similar effect may render networks of contract terms insignificant—a relatively unimportant single term in a form contract won’t affect the decisions of the contracting parties to adopt
the form. Network effects may exist in these component markets, but they are unlikely to be significant from a market perspective.517
Of course, numerous network effects will fall between these two extremes. Networks may
occupy only certain product market segments, or only certain geographic areas. As a general rule,
the significance of network effects positively correlates with their size in relation to product and
geographic markets. Even a “strong” network effect involving automobile fuel doors is unlikely to
be terribly important, for the simple reason that it will not affect purchasing decisions at the full
product level and therefore won’t determine market structure. By contrast, in industries in which
517.
Klausner makes a related point. See Klausner, supra note 5, at 814 (“for some contract terms, even if
network benefits are significant over some range of network scale, they reach a peak at some point”).
[Vol.
YEAR]
SHORT TITLE
187
the network defines the market, even weaker convenience-related effects can play a significant role
in determining market structure.
E. Importance of Existing Legal Rules
Network effects (strong or weak, big or small) do not occur in a vacuum. They are the product of economic, physical, and legal constraints that make up the market environment.518 Courts
evaluating legal claims of network effects should place particular emphasis on one of those sets of
constraints—the legal rules that contribute to network effects. Where the law contributes to, or
indeed causes, the market conditions that create network economic effects, courts should pay special attention to the policies behind such legal rules.
In many cases, of course, law is unrelated to or at most entirely incidental to network effects.
Even if one accepts the story of the QWERTY keyboard, for example, the standardization of that
keyboard occurred entirely without the intervention of the law. No law required people to use the
QWERTY keyboard, or even the same keyboard everyone else used. Nor did any legal rule (such
as an intellectual property right) restrain the production of keyboard formats by any private company. The law does not bind courts either to accept these alleged effects or attempt to change them.
We characterize this as a regime where the legal constraints on judicial action are low, but the
power of courts to act may be low as well.
Law may play a more significant role in network effects involving Klausner’s networks of
contracts.519 In theory, judicial decisions may increase the value of corporate governance terms by
clarifying their meaning and thus reducing the risk associated with using the term.
To the
518.
On this interrelationship, see Lessig, Constitution in Cyberspace, supra note 313, at 870.
519.
See Klausner, supra note 5, at 764-65; see supra notes 464-480 and accompanying text for a detailed
discussion.
++188
CALIFORNIA LAW REVIEW
(generally modest) extent that interpretive network effects may exist, courts function as the principal interpreters and possess a unique power to affect the strength of the network. At the same
time, legal constraints in the realm of contract exceed those in the QWERTY context: judicial action regarding standard contract terms influences the development of contract law in general and is
subject to the general limits of statutory authorization (or at least the threat of later legislative repudiation).520 To maximize clarity, courts should adhere to precedent to the extent possible and fully
articulate the reasons for their decisions, especially where they modify the law.521 In interpreting
terms such as "reasonable in relation to the threat posed," courts should do their best to identify
both the variables they deem important to a given decision and limitations on their rulings. Klausner’s regime is thus one of moderate judicial power, but moderate constraint as well.522
520.
The Delaware Supreme Court’s decision in Smith v. Van Gorkham, 488 A.2d 858 (Del. 1985), for
example, prompted the Delaware legislature to supplement the protections from liability avaialable to corporate
officers and directors. Forty-one other states soon followed suit. See Gilson & Black, supra note 401, at 1057.
521.
Kahan and Klausner also suggest that courts should adopt uniform interpretation of boilerplate terms
and give weight to the “legislative history” of the contractual equivalent of standard-setting organizations, such as the
American Bar Foundation’s Commentaries to its Model Debenture Indenture.
See Kahan & Klausner,
Standardization, supra note 61, at 762-66. These points have been recognized by several courts, and we agree with
them. See, e.g., Leverso v. Southtrust Bank of Al., Nat’l Assoc., 18 F.3d 1527, 1534 (11th Cir. 1994) (“the
meaning of boilerplate provisions is a matter of law and must be given a consistent, uniform interpretation”);
Sharon Steel Corp. v. Chase Manhattan Bank, 691 F.2d 1039, 1048 (2d Cir. 1982) (boilerplate provision, as
distinct from contractual provisions, “must be given a consistent, uniform interpretation”).
522.
Network theory does highlight the importance of judicial consideration of potential differences in the
economic world in which different firms operate and the differing market forces they face. Such an evaluation would
[Vol.
YEAR]
SHORT TITLE
189
Intellectual property law demonstrates an even greater relationship between legal doctrine and
network effects. In a fundamental sense, the structure of intellectual property law determines the
strength of network effects in a number of markets. Certainly, proprietary ownership of de facto
standards in the computer industry is an artifact of intellectual property law. Without intellectual
property law, Microsoft would not have a monopoly on PC operating systems, because every program they released would be rapidly and widely copied. This would solve many of the problems
associated with network effects: competition to sell the program would drive the price toward
marginal cost (near $0), expanding the number of people who used the standard software toward
the efficient level. Further, the ability of numerous competitors to improve on the Microsoft OS
with impunity would encourage migration within the standard to newer versions. Of course, there
are good policy reasons to support continued intellectual property protection for software, even
when it is a de facto standard. The existence of intellectual property protection is based on the concern that, without a means of appropriating at least some market power, no one will have an incentive to create new software.523 Intellectual property protection represents a considered judgment
that at least on average, the social value contributed by innovation, spurred by the incentive of intellectual property, outweighs the competitive harm caused by proprietary ownership of those innovations.
That Microsoft's proprietary network standard depends on intellectual property law means that
the law has more power over the network effects in this regime. Of course, significant legal con-
hopefully avoid transforming a general term that might otherwise provide room for tailoring into a majoritarian
default simply by virtue of judicial opinions that fail to account for differences among firms.
523.
See Lemley & O’Brien, supra note 38, at 830-31; McGowan, supra note 10, at 775-77; Pamela
Samuelson et al, A Manifesto Concerning the Legal Protection of Computer Programs, 94 Colum. L. Rev. 2308
(1994).
++190
CALIFORNIA LAW REVIEW
straints operate here as well. Intellectual property law is a creature of statute, and the statutes do
not generally permit the courts to withdraw intellectual property protection on an as-needed basis.
Further, some recent “takings” cases suggest that the Constitution may limit judicial discretion to
withdraw intellectual property protection.524 Nonetheless, intellectual property law does retain judicial flexibility in a number of areas, notably where the scope of intellectual property rights has not
been settled or where copyright’s fair use doctrine allows tailoring of protection to the circumstances. This flexibility means that courts can sometimes act within the confines of the law to
counter negative network effects, for example by favoring open standards or compelling interoperability. Thus, we would characterize the intellectual property regime as one of high power and
moderate constraint.
Finally, legal rules themselves sometimes compel the choice of a standard. One example is the
U.S. rule that autos must drive on the right side of the road. This standard results directly from
legal action and enforcement.525 Other examples include those limited cases in which the govern-
524.
See, e.g., Ruckelshaus v. Monsanto Co., 467 U.S. 986, 1016 (1984) (government statute compelling
disclosure of corporation’s product ingredients was a taking that required compensation). The Supreme Court’s 1996
decision in Seminole Tribe stripped Congress of the power to hold states liable for patent and copyright infringement
under the Commerce Clause. See Seminole Tribe of Florida v. Florida, 116 S. Ct. 1114, 1127 (1996). However,
in the wake of Seminole Tribe, some courts have held that patents and copyrights are “property” within the meaning
of the Fourteenth Amendment, and thus that Congress has the power to legislate regarding them in a way that is
binding on states. See, e.g., Genentech v. Regents of the University of California, 939 F. Supp. 639, 643 (S.D.
Ind. 1996); College Savings Bank v. Florida Prepaid Postsecondary Education Expense Board, 948 F. Supp. 400,
422-23 (D.N.J. 1997). If this approach is upheld, it should logically open the door to the argument that any
government interference with intellectual property rights is a “taking” of that property.
525.
Of course, the standard might have arisen anyway as a result of market forces.
[Vol.
YEAR]
SHORT TITLE
191
ment itself mandates a choice between competing private network standards, as in the case of highdefinition television (HDTV).526 Here, the law wields a high degree of power over the standard;
whether a product may be sold at all may depend on what the law says. However, the legal constraints on judicial action are normally also high, unless the legal rule has been established by or
delegated to the judiciary.
Judicial power over network effects, like the size and significance of those effects, is merely
one factor affecting the propriety of legal action to deal with network effects. In some circumstances, like the elimination of intellectual property rights, judicial action, though possible, may not
be appropriate for policy reasons. In others, judicial action may be a part of the effect itself. This
insight merely reinforces the importance of clarity in thinking through cases and drafting opinions.
In addition to ability to act and likely efficacy of action, courts must also consider the nature of the
legal action that must be taken. We consider this problem in the next section.
F. Nature of the Relief Sought
Finally, the nature of the judicial relief sought as a result of network effects will play a significant role in how courts evaluate and treat such effects. The judiciary should fashion relief in light
of the following considerations.
First, the mere existence of network effects is not outcome-determinative, as their presence
could be offered either to excuse conduct or to penalize it.527 As a general matter, network-effectsas-excuse arguments require less intrusive judicial action than those that would attempt to punish or
undo network effects. Excuse-based arguments do not require courts to attempt to alter or counter-
526.
See supra notes 271-274 and accompanying text (discussing government standard setting in HDTV).
527
See, for example, the arguments in the antitrust context. See supra notes 69-190 and accompanying
text.
++192
CALIFORNIA LAW REVIEW
act existing economic effects, merely to acknowledge their existence.528 In at least one sense, therefore, excuse-based arguments raise fewer problems than arguments requiring courts to take more
affirmative action. Even here, though, the problem of remedy poses complex questions, because
the same network argument may be used to seek an array of different remedies. While courts
might (and likely should) consider network effects a good reason not to attempt to fragment a standard (one level of “exculpation”), the sheer presence of such effects will not necessarily weigh
against the use of alternative legal rules to open the standard or, from a legislative perspective, to
regulate the prices that may be charged for its use. Ultimately, none of these alternatives may be
desirable,529 but the manner and degree to which network theory serves as an excuse will differ
with respect to each of them.
Second, even where the totality of circumstances requires some affirmative action, the nature
of the relationship between the network effects and existing legal rules may hamper, or at least
complicate, that action. In certain cases, where the network effects result from existing governmental support, withdrawing or conditioning that support may alter those effects. Intellectual property presents such a (comparatively) “easy” case.
Even if withdrawing intellectual property
entitlements to network standards proves impossible or undesirable,530 courts may effectively condition those entitlements (for example, by permitting reverse engineering or requiring group standards to remain open) in a way that ameliorates their negative effects. By contrast, a court will
encounter much more difficulty attempting to “undo” network effects established by economic
528.
Of course, excuse-based arguments may require greater alteration of legal rules in any given case,
depending on the way in which those rules are written.
529.
warranted.
And we can safely say that calls are made for such remedies more often than they are actually
[Vol.
YEAR]
SHORT TITLE
193
forces, as with some proposed antitrust solutions to network-based problems.531 Such a remedy,
where a court attempts to remake a market, not only necessarily requires more fact-specific and
potentially ongoing judicial involvement, but raises the potential for unintended consequences as
well.532 We do not suggest that such remedial measures are never appropriate, but their intrusive
nature certainly recommends caution.
Third, courts must consider how easily the claimed relief fits within the existing body of legal
doctrine at issue. This is in part a matter of judicial experience in taking economic circumstances
into account. Antitrust law has more thoroughly integrated economic theory and evidence, than,
say, the law of racial discrimination.533 Antitrust precedent more thoroughly integrate economic
analysis than is the case in many other fields, providing courts with a relatively better foundation
for integrating economic variables into the law. Similarly, legal arguments based on network effects will fare better if they can point to analogous use of network-based arguments in related
fields.
Courts must also consider the flexibility of particular legal doctrines. Network effects often
represent exceptional cases in the law, and accommodating them in all disciplines may have a distorting effect on the legal rule in “normal” cases. This is particularly true where the law adopts an
530.
See supra notes 233-236 and accompanying text (discussing the problems with such an approach).
531.
See supra notes 193-219 and 248-266 and accompanying text (discussing such approaches).
532.
Strict antitrust limitations on horizontal and vertical mergers in the 1950s and 1960s, for example,
may have contributed to the conglomerate merger phenomenon, and thus the likely inefficiencies associated with
such firms. See Gilson & Black, supra note 401, at 329, 339-346.
533.
See Herbert Hovenkamp, The Areeda-Turner Treatise in Antitrust Analysis, 41 Antitrust Bull. 815,
821 (1996) (describing antitrust as “an extreme case of applied economics”).
++194
CALIFORNIA LAW REVIEW
intentionally inflexible rule in an effort to reduce uncertainty.534 In certain circumstances, taking a
long-term view of a particular body of law may cause courts to ignore network effects that operate
only in a few cases, in order to avoid opening the floodgates to a variety of frivolous claims.535
This argument can be taken too far: legal rules so inflexible that they do not reflect reality are generally undesirable. But as a general matter, the less a particular legal doctrine must adapt to take
account of network effects, the better for all concerned.
This point brings up a final, related one: the desirability of minimizing distortion applies
among legal disciplines as well as within them. Courts should, where possible, eliminate a pernicious network effect by applying an existing legal doctrine, or fashioning a narrowly focused
holding, rather than making wholesale adaptations to a different legal discipline. We refer to this
as the principle of “minimal invasiveness.” This principle has important implications for judicial
choice of remedy once a problematic network effect has been identified. Generally, courts dealing
with a network-related problem on a narrow, fact-specific basis will create fewer distortions on
cases not involving network effects.536
534.
See id. (arguing that economic analysis in antitrust should sacrifice some accuracy for the sake of
simplicity, because “the more accurate measure is likely to invite complexity and, as a result, error”).
535.
Various commentators have made such arguments in antitrust law, contending that legal conclusions
such as the essential facilities doctrine or the intrabrand market, acknowledged in Eastman Kodak v. Image Technical
Servs., 504 U.S. 451, 501 (1992), even if technically correct, lead to enormous numbers of unsupported claims.
See, e.g., Areeda, Epithet, supra note 174, at 841; McGowan, supra note 10, at 804-06; Carl Shapiro, Aftermarkets
and Consumer Welfare: Making Sense of Kodak, 63 Antitrust L.J. 483, 485 (1995).
536.
Judge Easterbrook has argued that in a complex field like antitrust, legislative rules are more likely
than judicial decisions to get it right. See Frank Easterbrook, Ignorance and Antitrust, in Thomas M. Jorde & David
J. Teece, Antitrust, Innovation and Competitiveness 119, 123-29 (1992). We are not entirely persuaded. Cetainly,
[Vol.
YEAR]
SHORT TITLE
195
For example, the claim that something must be done about Microsoft's economic power may
potentially implicate antitrust, copyright, or contract law. If a court concluded that Microsoft had
obtained its position as the standard operating system by announcing an open architecture in order
to induce other firms to produce complementary software, and then closed the standard, a court
would then have to select the best legal tools to address the problem. Antitrust, which is keyed to
economic facts such as market share, provides relatively blunt remedies. Copyright, which has the
analytical tools that more closely fit the economic facts behind Microsoft's success, would provide
a relatively narrower approach and would avoid potential unintended consequences from an opinion keyed simply to market share. Even within copyright, as the First Circuit's opinion in Lotus v.
Borland demonstrates, a court could choose either to address the copyrightability of a type of software or to attempt to fashion more particularized remedies under the fair use doctrine.537
Still, although one could conceive of a fair use opinion tied in some way to the original pronouncement that created the problem in the first place, it is a bit of a stretch. A court might instead
look to contract law or estoppel, seeking to key the legal obligations imposed by the court to the
original pronouncements that induced third-party investment in compatible goods. Of the available
approaches, this last one would avoid potential negative consequences associated with a broad antitrust ruling limiting intellectual property rights, avoid problems associated with excluding certain
legislative rules might sometimes be tailored enough to be thought “minimally invasive”—consider Klausner’s
menus of contracts, for example. But by and large, legislative solutions are both more difficult and more general
than judicial decisions.
537.
See Lotus Dev. Corp. v. Borland Int’l, 49 F.3d 807, 821-22 (1st Cir. 1995), aff’d 516 U.S. 233
(1996), (Boudin, J., concurring) (exploring some of these choices); Lemley, supra note 201, at 1079-81 (suggesting
that Lotus should have been decided on more narrow grounds than in fact it was); McGowan, supra note 10, at 848
n.310 (suggesting Judge Boudin’s fair use approach was preferable to the majority’s).
++196
CALIFORNIA LAW REVIEW
portions of a product from copyright protection, and allow firms to choose ex ante the level of ongoing obligations they would assume. Such a rule might also have a beneficial information-forcing
function: if a firm producing a computer operating system secretly intended to leave its architecture
open only long enough to develop the critical mass of application programs necessary to help it win
the standards competition for operating systems, then it would have to reserve the right to close its
architecture in the future, potentially prompting third party firms to bargain for rights they might
not otherwise think they needed.538
Thus, we suggest that in particular cases, existing doctrines of contract law or equitable estoppel may serve as a better mechanism to compel open access to a standard-setting organization than
the essential facilities doctrine in antitrust;539 that tailored applications of the fair use doctrine in
copyright are a better way to alleviate problems caused by private ownership of network standards
than revocation of the copyright itself;540 and that compelling equal access to a network standard on
nondiscriminatory terms is normally preferable to government standard-setting.541 Further, some
network problems may correct themselves, either because the market itself will gravitate toward an
open standard, or because the ability of consumers to use multiple standards will defeat attempts at
exclusivity.542 In these latter cases, the “minimally invasive” alternative requires courts to do nothing at all, except ensure that no one deliberately impedes the market process.
538.
See McGowan, supra note 10, at 836-41 (elaborating on this approach).
539.
Such doctrines will not work in all cases, however. For some reasons why, see Lemley, Internet
Standardization, supra note 35, at 1090-92.
540.
See supra notes 231-247 and accompanying text (discussing these alternatives).
541.
See supra notes 273-284 and accompanying text (discussing these alternatives).
542.
Matutes and Regibeau demonstrate that under certain market conditions, notably those involving
consumer assembly of systems from component products, the equilibrium state involves full compatibility. See
[Vol.
YEAR]
SHORT TITLE
197
IV
Conclusions
Network effects theory provides an important example of the process by which the law learns
from and adapts to theories from other disciplines. Network theories are potentially important to
many legal fields and, because in some circumstances they challenge the classical economic presumption of declining returns to scale, may have far-reaching consequences for the development of
the law. The process of adaptation will be difficult, however, because of the variety of arguments
network theory can support and the differing, indeed opposing, conclusions to which such arguments lead. Courts confronted with arguments based on network theories should first ascertain the
nature of the claim asserted. We believe network arguments can best be categorized along a continuum, with actual networks—in effect communications systems—the strongest, and virtual networks— frequently involving interfaces between vertically related goods—providing a range of
examples of differing strength. At the other end of the continuum are a variety of phenomena in
which provision of a good or service positively relates to some level of scale, but in which the
scale economies themselves rather than interactions among users of the good creates the value.
After assessing the nature of the claim, courts should examine the market evidence advanced
in support of the claim. In doing so, courts must be careful not to infer the presence of network
effects solely from a coincidence of a large number of consumers of a particular good and significant economies of scale. Courts should focus on whether factors usually associated with scale
economies, most particularly declining average costs, or interactions among users of the good,
Carmen Matutes & Pierre Regibeau, “Mix and Match”: Product Compatibility Without Network Externalities, 19
RAND J. Econ. 221, [recommend pincite] (1988). To the extent this is true, legal efforts to force compatibility
in such market may prove unnecessary.
++198
CALIFORNIA LAW REVIEW
which provides stronger evidence of a potential network effect, generate the returns at issue. They
should also look for indications that the network effect does (or does not) play a significant role in
purchasing decisions, and identify what if any problems this causes.
If the network effects are real, significant, and negative, and if they cannot be solved within
the market, courts should then consider which legal doctrines offer the appropriate remedy. To the
extent that precedent permits, they should seek remedies from within the field of law that best takes
into account both the nature of the conduct complained of and whatever level of network effects
exist. Within that field, courts should seek to pursue the course that provides an adequate remedy
with the least possible disruption to other legal doctrines.
This analysis is complex and fact-specific. Some find this a flaw, suggesting that courts require simple rules rather than complex standards to guide their decision-making.543 We see it rather
as a virtue. Whatever the value of bright-line rules in circumstances in which courts know (or
543.
See, e.g, Gregory S. Crespi, Does the Chicago School Need to Expand It’s Curriculum?, 22 L. &
Soc. Inquiry 149, 154 (1997) (warning against the danger of overcomplicating economic analysis of law). But see
Thomas S. Ulen, Professor Crespi on Chicago, 22 L. & Soc. Inquiry 191, 197-98 (1997) (criticizing this approach).
The literature on what might be called the “rules vs. standards” debate is voluminous. See generally Isaac Ehrlich &
Richard A. Posner, An Economic Analysis of Legal Rulemaking, 3 J. Legal Stud. 257 (1974); Louis Kaplow,
Rules versus Standards: An Economic Analysis, 42 Duke L.J. 557 (1992). Besides the obvious parameters of
administrative cost and accuracy of result, more recent scholarship has suggested that adopting ambiguous standards
may actually have positive effects on bargaining. See, e.g., Ian Ayres & Eric Talley, Solomonic Bargaining:
Dividing a Legal Entitlement to Facilitate Coasean Trade, 104 Yale L.J. 1027 (1995); Dan L. Burk, Muddy Rules
for Cyberspace (working paper 1997); Jason Scott Johnston, Bargaining Under Rules Versus Standards, 11 J.L.
Econ. & Org. 256 (1995). Cf. Gillian K. Hadfield, Judicial Competence and the Interpretation of Incomplete
Contracts, 23 J. Legal Stud. 159 (1994) [recommend parenthetical].
[Vol.
YEAR]
SHORT TITLE
199
think they know) how to characterize conduct, the error costs of such categorization in network
markets today are simply too great. We do not know nearly as much as we would like about how
networks work and what they mean for market structure; what we do know suggests that network
“effects” may have either positive or negative ramifications depending on a whole host of factors.
It is sometimes efficient to give up accuracy for certainty and ease of administration. But here, the
accuracy we will lose will be significant, and the gain in administrative efficiency uncertain.
More generally, the example of network effects suggests some things about the role that economic theory (and by extension, other theories external to law) play in the judicial enterprise. All
law is purposive. Decisionmakers craft the statutory and common law with some end in mind.
Economic theory can generate predictions about how different legal rules will affect the state of the
world and, to that extent, provide important information to the legal decision-making process. If
the law wishes to induce the production of intellectual property, for example, rate of return analysis
should play an important role in any sensible intellectual property regime. However, economic
theory remains, or at least should remain, generally agnostic on normative matters: it cannot tell us
whether encouraging the production of intellectual property is something we should wish to do.
There is a difference between law and economics, an estimable discipline, and law as economics,
an unrealistic construct.
That said, courts and legislatures should obtain the best information available, and for many
fields of law, economic theory and data will prove useful to courts. To the degree the courts themselves may have inherent expertise in the legal field influenced by economic theory, such as certification of a plaintiff class in securities litigation, the process of adaptation is relatively safe and even
relatively new economic theories may be adapted quite successfully. The courts themselves possess the expertise to evaluate the process of adaptation and its potential consequences.
++200
CALIFORNIA LAW REVIEW
As we move into fields of substantive law, however, the picture becomes murkier as the usurpation of the federal antifraud laws by the efficient markets hypothesis also shows. In general,
legal disciplines inherently economic in character, such as antitrust, absorb new theories more easily than other fields. Assuming the purpose of the law is reasonably settled, and that the courts
have properly examined claims based on network effects, courts and legislatures should have a
better idea whether to accept or reject an argument based on network theory than in fields in which
economic theory has played a smaller role.
In short, these are not easy questions, and there are no categorical or easy answers. Network
economic theories promise to help courts refine legal rules to take account of market structures that
do not fit the classical model. But to take advantage of these theories, courts must not blindly accept any argument made under the rubric of “network eternality.” Rather, they must become adept
at separating the wheat from the chaff. We have tried to offer some first steps in that evaluative
process.
[Vol.